]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/vfs/vfs_subr.c
xnu-1699.22.73.tar.gz
[apple/xnu.git] / bsd / vfs / vfs_subr.c
index 3b10114cb52f4fc41442e4645fcd3536c5eb63af..462fbef7934d10bc99be6468b0c27b9b00d5ae47 100644 (file)
 
 #include <mach/mach_types.h>
 #include <mach/memory_object_types.h>
 
 #include <mach/mach_types.h>
 #include <mach/memory_object_types.h>
+#include <mach/memory_object_control.h>
 
 #include <kern/kalloc.h>       /* kalloc()/kfree() */
 #include <kern/clock.h>                /* delay_for_interval() */
 #include <libkern/OSAtomic.h>  /* OSAddAtomic() */
 
 
 
 #include <kern/kalloc.h>       /* kalloc()/kfree() */
 #include <kern/clock.h>                /* delay_for_interval() */
 #include <libkern/OSAtomic.h>  /* OSAddAtomic() */
 
 
+#ifdef JOE_DEBUG
+#include <libkern/OSDebug.h>
+#endif
+
 #include <vm/vm_protos.h>      /* vnode_pager_vrele() */
 
 #if CONFIG_MACF
 #include <vm/vm_protos.h>      /* vnode_pager_vrele() */
 
 #if CONFIG_MACF
 extern lck_grp_t *vnode_lck_grp;
 extern lck_attr_t *vnode_lck_attr;
 
 extern lck_grp_t *vnode_lck_grp;
 extern lck_attr_t *vnode_lck_attr;
 
+#if CONFIG_TRIGGERS
+extern lck_grp_t *trigger_vnode_lck_grp;
+extern lck_attr_t *trigger_vnode_lck_attr;
+#endif
 
 extern lck_mtx_t * mnt_list_mtx_lock;
 
 
 extern lck_mtx_t * mnt_list_mtx_lock;
 
@@ -145,6 +154,16 @@ int        vttoif_tab[9] = {
        S_IFSOCK, S_IFIFO, S_IFMT,
 };
 
        S_IFSOCK, S_IFIFO, S_IFMT,
 };
 
+
+/* XXX These should be in a BSD accessible Mach header, but aren't. */
+extern void             memory_object_mark_used(
+       memory_object_control_t         control);
+
+extern void             memory_object_mark_unused(
+       memory_object_control_t         control,
+       boolean_t                       rage);
+
+
 /* XXX next protptype should be from <nfs/nfs.h> */
 extern int       nfs_vinvalbuf(vnode_t, int, vfs_context_t, int);
 
 /* XXX next protptype should be from <nfs/nfs.h> */
 extern int       nfs_vinvalbuf(vnode_t, int, vfs_context_t, int);
 
@@ -173,7 +192,6 @@ static void vclean(vnode_t vp, int flag);
 static void vnode_reclaim_internal(vnode_t, int, int, int);
 
 static void vnode_dropiocount (vnode_t);
 static void vnode_reclaim_internal(vnode_t, int, int, int);
 
 static void vnode_dropiocount (vnode_t);
-static errno_t vnode_getiocount(vnode_t vp, unsigned int vid, int vflags);
 
 static vnode_t checkalias(vnode_t vp, dev_t nvp_rdev);
 static int  vnode_reload(vnode_t);
 
 static vnode_t checkalias(vnode_t vp, dev_t nvp_rdev);
 static int  vnode_reload(vnode_t);
@@ -188,6 +206,9 @@ static int vnode_iterate_prepare(mount_t);
 static int vnode_iterate_reloadq(mount_t);
 static void vnode_iterate_clear(mount_t);
 static mount_t vfs_getvfs_locked(fsid_t *);
 static int vnode_iterate_reloadq(mount_t);
 static void vnode_iterate_clear(mount_t);
 static mount_t vfs_getvfs_locked(fsid_t *);
+static int vn_create_reg(vnode_t dvp, vnode_t *vpp, struct nameidata *ndp,
+               struct vnode_attr *vap, uint32_t flags, int fmode, uint32_t *statusp, vfs_context_t ctx);
+static int vnode_authattr_new_internal(vnode_t dvp, struct vnode_attr *vap, int noauth, uint32_t *defaulted_fieldsp, vfs_context_t ctx);
 
 errno_t rmdir_remove_orphaned_appleDouble(vnode_t, vfs_context_t, int *); 
 
 
 errno_t rmdir_remove_orphaned_appleDouble(vnode_t, vfs_context_t, int *); 
 
@@ -195,6 +216,11 @@ errno_t rmdir_remove_orphaned_appleDouble(vnode_t, vfs_context_t, int *);
 static void record_vp(vnode_t vp, int count);
 #endif
 
 static void record_vp(vnode_t vp, int count);
 #endif
 
+#if CONFIG_TRIGGERS
+static int vnode_resolver_create(mount_t, vnode_t, struct vnode_trigger_param *, boolean_t external);
+static void vnode_resolver_detach(vnode_t);
+#endif
+
 TAILQ_HEAD(freelst, vnode) vnode_free_list;    /* vnode free list */
 TAILQ_HEAD(deadlst, vnode) vnode_dead_list;    /* vnode dead list */
 
 TAILQ_HEAD(freelst, vnode) vnode_free_list;    /* vnode free list */
 TAILQ_HEAD(deadlst, vnode) vnode_dead_list;    /* vnode dead list */
 
@@ -370,29 +396,27 @@ void
 vnode_writedone(vnode_t vp)
 {
        if (vp) {
 vnode_writedone(vnode_t vp)
 {
        if (vp) {
-               OSAddAtomic(-1, &vp->v_numoutput);
+               int need_wakeup = 0;
 
 
-               if (vp->v_numoutput <= 1) {
-                       int need_wakeup = 0;
+               OSAddAtomic(-1, &vp->v_numoutput);
 
 
-                       vnode_lock_spin(vp);
+               vnode_lock_spin(vp);
 
 
-                       if (vp->v_numoutput < 0)
-                               panic("vnode_writedone: numoutput < 0");
+               if (vp->v_numoutput < 0)
+                       panic("vnode_writedone: numoutput < 0");
 
 
-                       if ((vp->v_flag & VTHROTTLED) && (vp->v_numoutput <= 1)) {
-                               vp->v_flag &= ~VTHROTTLED;
-                               need_wakeup = 1;
-                       }
-                       if ((vp->v_flag & VBWAIT) && (vp->v_numoutput == 0)) {
-                               vp->v_flag &= ~VBWAIT;
-                               need_wakeup = 1;
-                       }
-                       vnode_unlock(vp);
-               
-                       if (need_wakeup)
-                               wakeup((caddr_t)&vp->v_numoutput);
+               if ((vp->v_flag & VTHROTTLED)) {
+                       vp->v_flag &= ~VTHROTTLED;
+                       need_wakeup = 1;
+               }
+               if ((vp->v_flag & VBWAIT) && (vp->v_numoutput == 0)) {
+                       vp->v_flag &= ~VBWAIT;
+                       need_wakeup = 1;
                }
                }
+               vnode_unlock(vp);
+               
+               if (need_wakeup)
+                       wakeup((caddr_t)&vp->v_numoutput);
        }
 }
 
        }
 }
 
@@ -781,6 +805,13 @@ mount_refdrain(mount_t mp)
        return(0);
 }
 
        return(0);
 }
 
+/* Tags the mount point as not supportine extended readdir for NFS exports */
+void 
+mount_set_noreaddirext(mount_t mp) {
+       mount_lock (mp);
+       mp->mnt_kern_flag |= MNTK_DENY_READDIREXT;
+       mount_unlock (mp);
+}
 
 /*
  * Mark a mount point as busy. Used to synchronize access and to delay
 
 /*
  * Mark a mount point as busy. Used to synchronize access and to delay
@@ -887,6 +918,8 @@ vfs_rootmountalloc_internal(struct vfstable *vfsp, const char *devname)
        mp->mnt_ioflags = 0;
        mp->mnt_realrootvp = NULLVP;
        mp->mnt_authcache_ttl = CACHED_LOOKUP_RIGHT_TTL;
        mp->mnt_ioflags = 0;
        mp->mnt_realrootvp = NULLVP;
        mp->mnt_authcache_ttl = CACHED_LOOKUP_RIGHT_TTL;
+       mp->mnt_throttle_mask = LOWPRI_MAX_NUM_DEV - 1;
+       mp->mnt_devbsdunit = 0;
 
        mount_lock_init(mp);
        (void)vfs_busy(mp, LK_NOWAIT);
 
        mount_lock_init(mp);
        (void)vfs_busy(mp, LK_NOWAIT);
@@ -1120,6 +1153,8 @@ vfs_getvfs_by_mntonname(char *path)
                if (!strncmp(mp->mnt_vfsstat.f_mntonname, path,
                                        sizeof(mp->mnt_vfsstat.f_mntonname))) {
                        retmp = mp;
                if (!strncmp(mp->mnt_vfsstat.f_mntonname, path,
                                        sizeof(mp->mnt_vfsstat.f_mntonname))) {
                        retmp = mp;
+                       if (mount_iterref(retmp, 1))
+                               retmp = NULL;
                        goto out;
                }
        }
                        goto out;
                }
        }
@@ -1358,6 +1393,7 @@ found_alias:
                nvp->v_rdev = nvp_rdev;
                nvp->v_specflags = 0;
                nvp->v_speclastr = -1;
                nvp->v_rdev = nvp_rdev;
                nvp->v_specflags = 0;
                nvp->v_speclastr = -1;
+               nvp->v_specinfo->si_opencount = 0;
 
                SPECHASH_LOCK();
                
 
                SPECHASH_LOCK();
                
@@ -1416,22 +1452,16 @@ int
 vget_internal(vnode_t vp, int vid, int vflags)
 {
        int error = 0;
 vget_internal(vnode_t vp, int vid, int vflags)
 {
        int error = 0;
-       int vpid;
 
        vnode_lock_spin(vp);
 
 
        vnode_lock_spin(vp);
 
-       if (vflags & VNODE_WITHID)
-               vpid = vid;
-       else
-               vpid = vp->v_id;    // save off the original v_id
-
        if ((vflags & VNODE_WRITEABLE) && (vp->v_writecount == 0))
                /*
                 * vnode to be returned only if it has writers opened 
                 */
                error = EINVAL;
        else
        if ((vflags & VNODE_WRITEABLE) && (vp->v_writecount == 0))
                /*
                 * vnode to be returned only if it has writers opened 
                 */
                error = EINVAL;
        else
-               error = vnode_getiocount(vp, vpid, vflags);
+               error = vnode_getiocount(vp, vid, vflags);
 
        vnode_unlock(vp);
 
 
        vnode_unlock(vp);
 
@@ -1446,7 +1476,7 @@ int
 vnode_ref(vnode_t vp)
 {
 
 vnode_ref(vnode_t vp)
 {
 
-        return (vnode_ref_ext(vp, 0));
+        return (vnode_ref_ext(vp, 0, 0));
 }
 
 /*
 }
 
 /*
@@ -1454,7 +1484,7 @@ vnode_ref(vnode_t vp)
  *             ENOENT                  No such file or directory [terminating]
  */
 int
  *             ENOENT                  No such file or directory [terminating]
  */
 int
-vnode_ref_ext(vnode_t vp, int fmode)
+vnode_ref_ext(vnode_t vp, int fmode, int flags)
 {
        int     error = 0;
 
 {
        int     error = 0;
 
@@ -1471,10 +1501,12 @@ vnode_ref_ext(vnode_t vp, int fmode)
        /*
         * if you are the owner of drain/termination, can acquire usecount
         */
        /*
         * if you are the owner of drain/termination, can acquire usecount
         */
-       if ((vp->v_lflag & (VL_DRAIN | VL_TERMINATE | VL_DEAD))) {
-               if (vp->v_owner != current_thread()) {
-                       error = ENOENT;
-                       goto out;
+       if ((flags & VNODE_REF_FORCE) == 0) {
+               if ((vp->v_lflag & (VL_DRAIN | VL_TERMINATE | VL_DEAD))) {
+                       if (vp->v_owner != current_thread()) {
+                               error = ENOENT;
+                               goto out;
+                       }
                }
        }
        vp->v_usecount++;
                }
        }
        vp->v_usecount++;
@@ -1507,6 +1539,13 @@ vnode_ref_ext(vnode_t vp, int fmode)
                        vnode_list_remove(vp);
                }
        }
                        vnode_list_remove(vp);
                }
        }
+       if (vp->v_usecount == 1 && vp->v_type == VREG && !(vp->v_flag & VSYSTEM)) {
+
+               if (vp->v_ubcinfo) {
+                       vnode_lock_convert(vp);
+                       memory_object_mark_used(vp->v_ubcinfo->ui_control);
+               }
+       }
 out:
        vnode_unlock(vp);
 
 out:
        vnode_unlock(vp);
 
@@ -1659,6 +1698,7 @@ vnode_rele_ext(vnode_t vp, int fmode, int dont_reenter)
 void
 vnode_rele_internal(vnode_t vp, int fmode, int dont_reenter, int locked)
 {
 void
 vnode_rele_internal(vnode_t vp, int fmode, int dont_reenter, int locked)
 {
+
        if ( !locked)
                vnode_lock_spin(vp);
 #if DIAGNOSTIC
        if ( !locked)
                vnode_lock_spin(vp);
 #if DIAGNOSTIC
@@ -1689,9 +1729,7 @@ vnode_rele_internal(vnode_t vp, int fmode, int dont_reenter, int locked)
                        vp->v_lflag |= VL_NEEDINACTIVE;
                        vp->v_flag  &= ~(VNOCACHE_DATA | VRAOFF | VOPENEVT);
                }
                        vp->v_lflag |= VL_NEEDINACTIVE;
                        vp->v_flag  &= ~(VNOCACHE_DATA | VRAOFF | VOPENEVT);
                }
-               if ( !locked)
-                       vnode_unlock(vp);
-               return;
+               goto done;
        }
        vp->v_flag  &= ~(VNOCACHE_DATA | VRAOFF | VOPENEVT);
 
        }
        vp->v_flag  &= ~(VNOCACHE_DATA | VRAOFF | VOPENEVT);
 
@@ -1709,9 +1747,8 @@ vnode_rele_internal(vnode_t vp, int fmode, int dont_reenter, int locked)
                        vp->v_flag |= VAGE;
                }
                vnode_list_add(vp);
                        vp->v_flag |= VAGE;
                }
                vnode_list_add(vp);
-               if ( !locked)
-                       vnode_unlock(vp);
-               return;
+
+               goto done;
        }
        /*
         * at this point both the iocount and usecount
        }
        /*
         * at this point both the iocount and usecount
@@ -1746,15 +1783,22 @@ vnode_rele_internal(vnode_t vp, int fmode, int dont_reenter, int locked)
                
                if (ut->uu_defer_reclaims) {
                        vp->v_defer_reclaimlist = ut->uu_vreclaims;
                
                if (ut->uu_defer_reclaims) {
                        vp->v_defer_reclaimlist = ut->uu_vreclaims;
-                               ut->uu_vreclaims = vp;
-                       goto defer_reclaim;
+                       ut->uu_vreclaims = vp;
+                       goto done;
                }
                vnode_lock_convert(vp);
                vnode_reclaim_internal(vp, 1, 1, 0);
        }
        vnode_dropiocount(vp);
        vnode_list_add(vp);
                }
                vnode_lock_convert(vp);
                vnode_reclaim_internal(vp, 1, 1, 0);
        }
        vnode_dropiocount(vp);
        vnode_list_add(vp);
-defer_reclaim:
+done:
+       if (vp->v_usecount == 0 && vp->v_type == VREG && !(vp->v_flag & VSYSTEM)) {
+
+               if (vp->v_ubcinfo) {
+                       vnode_lock_convert(vp);
+                       memory_object_mark_unused(vp->v_ubcinfo->ui_control, (vp->v_flag & VRAGE) == VRAGE);
+               }
+       }
        if ( !locked)
                vnode_unlock(vp);
        return;
        if ( !locked)
                vnode_unlock(vp);
        return;
@@ -2020,13 +2064,13 @@ vclean(vnode_t vp, int flags)
 #endif
                {
                        VNOP_FSYNC(vp, MNT_WAIT, ctx);
 #endif
                {
                        VNOP_FSYNC(vp, MNT_WAIT, ctx);
-                       buf_invalidateblks(vp, BUF_WRITE_DATA, 0, 0);
+                       buf_invalidateblks(vp, BUF_WRITE_DATA | BUF_INVALIDATE_LOCKED, 0, 0);
                }
                if (UBCINFOEXISTS(vp))
                        /*
                         * Clean the pages in VM.
                         */
                }
                if (UBCINFOEXISTS(vp))
                        /*
                         * Clean the pages in VM.
                         */
-                       (void)ubc_sync_range(vp, (off_t)0, ubc_getsize(vp), UBC_PUSHALL);
+                       (void)ubc_msync(vp, (off_t)0, ubc_getsize(vp), NULL, UBC_PUSHALL | UBC_INVALIDATE | UBC_SYNC);
        }
        if (active || need_inactive) 
                VNOP_INACTIVE(vp, ctx);
        }
        if (active || need_inactive) 
                VNOP_INACTIVE(vp, ctx);
@@ -2039,18 +2083,7 @@ vclean(vnode_t vp, int flags)
                if (vnode_isshadow(vp)) {
                        vnode_relenamedstream(pvp, vp, ctx);
                }
                if (vnode_isshadow(vp)) {
                        vnode_relenamedstream(pvp, vp, ctx);
                }
-
-               /*
-                * Because vclean calls VNOP_INACTIVE prior to calling vnode_relenamedstream, we may not have 
-                * torn down and/or deleted the shadow file yet.  On HFS, if the shadow file is sufficiently large 
-                * and occupies a large number of extents, the deletion will be deferred until VNOP_INACTIVE 
-                * and the file treated like an open-unlinked.  To rectify this, call VNOP_INACTIVE again
-                * explicitly to force its removal.
-                */
-               if (vnode_isshadow(vp)) {
-                       VNOP_INACTIVE(vp, ctx);
-               }
-
+               
                /* 
                 * No more streams associated with the parent.  We
                 * have a ref on it, so its identity is stable.
                /* 
                 * No more streams associated with the parent.  We
                 * have a ref on it, so its identity is stable.
@@ -2072,6 +2105,14 @@ vclean(vnode_t vp, int flags)
         */
        ubc_destroy_named(vp);
 
         */
        ubc_destroy_named(vp);
 
+#if CONFIG_TRIGGERS
+       /*
+        * cleanup trigger info from vnode (if any)
+        */
+       if (vp->v_resolve)
+               vnode_resolver_detach(vp);
+#endif
+
        /*
         * Reclaim the vnode.
         */
        /*
         * Reclaim the vnode.
         */
@@ -2301,7 +2342,7 @@ vcount(vnode_t vp)
 
 loop:
        if (!vnode_isaliased(vp))
 
 loop:
        if (!vnode_isaliased(vp))
-               return (vp->v_usecount - vp->v_kusecount);
+               return (vp->v_specinfo->si_opencount);
        count = 0;
 
        SPECHASH_LOCK();
        count = 0;
 
        SPECHASH_LOCK();
@@ -2332,7 +2373,7 @@ loop:
                                vnode_unlock(vq);
                                goto loop;
                        }
                                vnode_unlock(vq);
                                goto loop;
                        }
-                       count += (vq->v_usecount - vq->v_kusecount);
+                       count += vq->v_specinfo->si_opencount;
                }
                vnode_unlock(vq);
 
                }
                vnode_unlock(vq);
 
@@ -2710,7 +2751,7 @@ sysctl_vnode
 }
 
 SYSCTL_PROC(_kern, KERN_VNODE, vnode,
 }
 
 SYSCTL_PROC(_kern, KERN_VNODE, vnode,
-               CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MASKED,
+               CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MASKED | CTLFLAG_LOCKED,
                0, 0, sysctl_vnode, "S,", "");
 
 
                0, 0, sysctl_vnode, "S,", "");
 
 
@@ -2804,6 +2845,8 @@ vnode_pager_vrele(vnode_t vp)
 
 #include <sys/disk.h>
 
 
 #include <sys/disk.h>
 
+u_int32_t rootunit = (u_int32_t)-1;
+
 errno_t
 vfs_init_io_attributes(vnode_t devvp, mount_t mp)
 {
 errno_t
 vfs_init_io_attributes(vnode_t devvp, mount_t mp)
 {
@@ -2824,24 +2867,25 @@ vfs_init_io_attributes(vnode_t devvp, mount_t mp)
        vfs_context_t ctx = vfs_context_current();
        int isssd = 0;
        int isvirtual = 0;
        vfs_context_t ctx = vfs_context_current();
        int isssd = 0;
        int isvirtual = 0;
+
+
+       VNOP_IOCTL(devvp, DKIOCGETTHROTTLEMASK, (caddr_t)&mp->mnt_throttle_mask, 0, NULL);
        /*
        /*
-        * determine if this mount point exists on the same device as the root
-        * partition... if so, then it comes under the hard throttle control
+        * as a reasonable approximation, only use the lowest bit of the mask
+        * to generate a disk unit number
         */
         */
-       int        thisunit = -1;
-       static int rootunit = -1;
+       mp->mnt_devbsdunit = num_trailing_0(mp->mnt_throttle_mask);
 
 
-       if (rootunit == -1) {
-               if (VNOP_IOCTL(rootvp, DKIOCGETBSDUNIT, (caddr_t)&rootunit, 0, ctx))
-                       rootunit = -1; 
-               else if (rootvp == devvp)
-                       mp->mnt_kern_flag |= MNTK_ROOTDEV;
-       }
-       if (devvp != rootvp && rootunit != -1) {
-               if (VNOP_IOCTL(devvp, DKIOCGETBSDUNIT, (caddr_t)&thisunit, 0, ctx) == 0) {
-                       if (thisunit == rootunit)
-                               mp->mnt_kern_flag |= MNTK_ROOTDEV;
-               }
+       if (devvp == rootvp)
+               rootunit = mp->mnt_devbsdunit;
+
+       if (mp->mnt_devbsdunit == rootunit) {
+               /*
+                * this mount point exists on the same device as the root
+                * partition, so it comes under the hard throttle control...
+                * this is true even for the root mount point itself
+                */
+               mp->mnt_kern_flag |= MNTK_ROOTDEV;
        }
        /*
         * force the spec device to re-cache
        }
        /*
         * force the spec device to re-cache
@@ -2875,7 +2919,6 @@ vfs_init_io_attributes(vnode_t devvp, mount_t mp)
                if (isssd)
                        mp->mnt_kern_flag |= MNTK_SSD;
        }
                if (isssd)
                        mp->mnt_kern_flag |= MNTK_SSD;
        }
-
        if ((error = VNOP_IOCTL(devvp, DKIOCGETFEATURES,
                                (caddr_t)&features, 0, ctx)))
                return (error);
        if ((error = VNOP_IOCTL(devvp, DKIOCGETFEATURES,
                                (caddr_t)&features, 0, ctx)))
                return (error);
@@ -3253,7 +3296,11 @@ sysctl_vfs_ctlbyfsid(__unused struct sysctl_oid *oidp, void *arg1, int arg2,
                        sfs.f_fsid = sp->f_fsid;
                        sfs.f_owner = sp->f_owner;
     
                        sfs.f_fsid = sp->f_fsid;
                        sfs.f_owner = sp->f_owner;
     
-                       strlcpy(sfs.f_fstypename, sp->f_fstypename, MFSNAMELEN);
+                       if (mp->mnt_kern_flag & MNTK_TYPENAME_OVERRIDE) {
+                               strlcpy(&sfs.f_fstypename[0], &mp->fstypename_override[0], MFSTYPENAMELEN);
+                       } else {
+                               strlcpy(sfs.f_fstypename, sp->f_fstypename, MFSNAMELEN);
+                       }
                        strlcpy(sfs.f_mntonname, sp->f_mntonname, MNAMELEN);
                        strlcpy(sfs.f_mntfromname, sp->f_mntfromname, MNAMELEN);
             
                        strlcpy(sfs.f_mntonname, sp->f_mntonname, MNAMELEN);
                        strlcpy(sfs.f_mntfromname, sp->f_mntfromname, MNAMELEN);
             
@@ -3307,7 +3354,11 @@ sysctl_vfs_ctlbyfsid(__unused struct sysctl_oid *oidp, void *arg1, int arg2,
                        sfs.f_fsid = sp->f_fsid;
                        sfs.f_owner = sp->f_owner;
     
                        sfs.f_fsid = sp->f_fsid;
                        sfs.f_owner = sp->f_owner;
     
-                       strlcpy(sfs.f_fstypename, sp->f_fstypename, MFSNAMELEN);
+                       if (mp->mnt_kern_flag & MNTK_TYPENAME_OVERRIDE) {
+                               strlcpy(&sfs.f_fstypename[0], &mp->fstypename_override[0], MFSTYPENAMELEN);
+                       } else {
+                               strlcpy(sfs.f_fstypename, sp->f_fstypename, MFSNAMELEN);
+                       }
                        strlcpy(sfs.f_mntonname, sp->f_mntonname, MNAMELEN);
                        strlcpy(sfs.f_mntfromname, sp->f_mntfromname, MNAMELEN);
             
                        strlcpy(sfs.f_mntonname, sp->f_mntonname, MNAMELEN);
                        strlcpy(sfs.f_mntfromname, sp->f_mntfromname, MNAMELEN);
             
@@ -3414,14 +3465,14 @@ sysctl_vfs_noremotehang(__unused struct sysctl_oid *oidp,
 }
 
 /* the vfs.generic. branch. */
 }
 
 /* the vfs.generic. branch. */
-SYSCTL_NODE(_vfs, VFS_GENERIC, generic, CTLFLAG_RW|CTLFLAG_LOCKED, NULL, "vfs generic hinge");
+SYSCTL_NODE(_vfs, VFS_GENERIC, generic, CTLFLAG_RW | CTLFLAG_LOCKED, NULL, "vfs generic hinge");
 /* retreive a list of mounted filesystem fsid_t */
 /* retreive a list of mounted filesystem fsid_t */
-SYSCTL_PROC(_vfs_generic, OID_AUTO, vfsidlist, CTLFLAG_RD,
+SYSCTL_PROC(_vfs_generic, OID_AUTO, vfsidlist, CTLFLAG_RD | CTLFLAG_LOCKED,
     NULL, 0, sysctl_vfs_vfslist, "S,fsid", "List of mounted filesystem ids");
 /* perform operations on filesystem via fsid_t */
     NULL, 0, sysctl_vfs_vfslist, "S,fsid", "List of mounted filesystem ids");
 /* perform operations on filesystem via fsid_t */
-SYSCTL_NODE(_vfs_generic, OID_AUTO, ctlbyfsid, CTLFLAG_RW|CTLFLAG_LOCKED,
+SYSCTL_NODE(_vfs_generic, OID_AUTO, ctlbyfsid, CTLFLAG_RW | CTLFLAG_LOCKED,
     sysctl_vfs_ctlbyfsid, "ctlbyfsid");
     sysctl_vfs_ctlbyfsid, "ctlbyfsid");
-SYSCTL_PROC(_vfs_generic, OID_AUTO, noremotehang, CTLFLAG_RW|CTLFLAG_ANYBODY,
+SYSCTL_PROC(_vfs_generic, OID_AUTO, noremotehang, CTLFLAG_RW | CTLFLAG_ANYBODY,
     NULL, 0, sysctl_vfs_noremotehang, "I", "noremotehang");
        
        
     NULL, 0, sysctl_vfs_noremotehang, "I", "noremotehang");
        
        
@@ -3448,18 +3499,18 @@ retry:
 
        vnode_list_lock();
 
 
        vnode_list_lock();
 
-       if ( !TAILQ_EMPTY(&vnode_dead_list)) {
-               /*
-                * Can always reuse a dead one
+       if ((numvnodes - deadvnodes) < desiredvnodes || force_alloc) {
+               if ( !TAILQ_EMPTY(&vnode_dead_list)) {
+                       /*
+                        * Can always reuse a dead one
+                        */
+                       vp = TAILQ_FIRST(&vnode_dead_list);
+                       goto steal_this_vp;
+               }
+               /*
+                * no dead vnodes available... if we're under
+                * the limit, we'll create a new vnode
                 */
                 */
-               vp = TAILQ_FIRST(&vnode_dead_list);
-               goto steal_this_vp;
-       }
-       /*
-        * no dead vnodes available... if we're under
-        * the limit, we'll create a new vnode
-        */
-       if (numvnodes < desiredvnodes || force_alloc) {
                numvnodes++;
                vnode_list_unlock();
 
                numvnodes++;
                vnode_list_unlock();
 
@@ -3493,17 +3544,22 @@ retry:
                        panic("new_vnode: vp (%p) on RAGE list not marked VLIST_RAGE", vp);
 
                    // if we're a dependency-capable process, skip vnodes that can
                        panic("new_vnode: vp (%p) on RAGE list not marked VLIST_RAGE", vp);
 
                    // if we're a dependency-capable process, skip vnodes that can
-                   // cause recycling deadlocks. (i.e. this process is diskimages
-                   // helper and the vnode is in a disk image).
-                   //
-                   if ((curproc->p_flag & P_DEPENDENCY_CAPABLE) == 0 || vp->v_mount == NULL || vp->v_mount->mnt_dependent_process == NULL) {
-                       break;
+                       // cause recycling deadlocks. (i.e. this process is diskimages
+                       // helper and the vnode is in a disk image).  Querying the
+                       // mnt_kern_flag for the mount's virtual device status
+                       // is safer than checking the mnt_dependent_process, which
+                       // may not be updated if there are multiple devnode layers 
+                       // in between the disk image and the final consumer.
+
+                   if ((curproc->p_flag & P_DEPENDENCY_CAPABLE) == 0 || vp->v_mount == NULL || 
+                                       (vp->v_mount->mnt_kern_flag & MNTK_VIRTUALDEV) == 0) {
+                               break;
                    }
 
                    // don't iterate more than MAX_WALK_COUNT vnodes to
                    // avoid keeping the vnode list lock held for too long.
                    if (walk_count++ > MAX_WALK_COUNT) {
                    }
 
                    // don't iterate more than MAX_WALK_COUNT vnodes to
                    // avoid keeping the vnode list lock held for too long.
                    if (walk_count++ > MAX_WALK_COUNT) {
-                       vp = NULL;
+                               vp = NULL;
                        break;
                    }
                }
                        break;
                    }
                }
@@ -3516,12 +3572,18 @@ retry:
                 */
                walk_count = 0;
                TAILQ_FOREACH(vp, &vnode_free_list, v_freelist) {
                 */
                walk_count = 0;
                TAILQ_FOREACH(vp, &vnode_free_list, v_freelist) {
-                   // if we're a dependency-capable process, skip vnodes that can
-                   // cause recycling deadlocks. (i.e. this process is diskimages
-                   // helper and the vnode is in a disk image)
-                   //
-                   if ((curproc->p_flag & P_DEPENDENCY_CAPABLE) == 0 || vp->v_mount == NULL || vp->v_mount->mnt_dependent_process == NULL) {
-                       break;
+
+                       // if we're a dependency-capable process, skip vnodes that can
+                       // cause recycling deadlocks. (i.e. this process is diskimages
+                       // helper and the vnode is in a disk image).  Querying the
+                       // mnt_kern_flag for the mount's virtual device status
+                       // is safer than checking the mnt_dependent_process, which
+                       // may not be updated if there are multiple devnode layers 
+                       // in between the disk image and the final consumer.
+
+                   if ((curproc->p_flag & P_DEPENDENCY_CAPABLE) == 0 || vp->v_mount == NULL || 
+                                       (vp->v_mount->mnt_kern_flag & MNTK_VIRTUALDEV) == 0) {
+                               break;
                    }
 
                    // don't iterate more than MAX_WALK_COUNT vnodes to
                    }
 
                    // don't iterate more than MAX_WALK_COUNT vnodes to
@@ -3572,7 +3634,7 @@ retry:
                 * Running out of vnodes tends to make a system unusable. Start killing
                 * processes that jetsam knows are killable.
                 */
                 * Running out of vnodes tends to make a system unusable. Start killing
                 * processes that jetsam knows are killable.
                 */
-               if (jetsam_kill_top_proc() < 0) {
+               if (jetsam_kill_top_proc(TRUE, kJetsamFlagsKilledVnodes) < 0) {
                        /*
                         * If jetsam can't find any more processes to kill and there
                         * still aren't any free vnodes, panic. Hopefully we'll get a
                        /*
                         * If jetsam can't find any more processes to kill and there
                         * still aren't any free vnodes, panic. Hopefully we'll get a
@@ -3754,10 +3816,27 @@ vnode_get_locked(struct vnode *vp)
        return (0);
 }
 
        return (0);
 }
 
+/*
+ * vnode_getwithvid() cuts in line in front of a vnode drain (that is,
+ * while the vnode is draining, but at no point after that) to prevent
+ * deadlocks when getting vnodes from filesystem hashes while holding
+ * resources that may prevent other iocounts from being released.
+ */
 int
 vnode_getwithvid(vnode_t vp, uint32_t vid)
 {
 int
 vnode_getwithvid(vnode_t vp, uint32_t vid)
 {
-        return(vget_internal(vp, vid, ( VNODE_NODEAD| VNODE_WITHID)));
+        return(vget_internal(vp, vid, ( VNODE_NODEAD | VNODE_WITHID | VNODE_DRAINO )));
+}
+
+/*
+ * vnode_getwithvid_drainok() is like vnode_getwithvid(), but *does* block behind a vnode
+ * drain; it exists for use in the VFS name cache, where we really do want to block behind
+ * vnode drain to prevent holding off an unmount.
+ */
+int
+vnode_getwithvid_drainok(vnode_t vp, uint32_t vid)
+{
+        return(vget_internal(vp, vid, ( VNODE_NODEAD | VNODE_WITHID )));
 }
 
 int
 }
 
 int
@@ -3801,7 +3880,7 @@ retry:
                vnode_dropiocount(vp);
                return(0);
        }
                vnode_dropiocount(vp);
                return(0);
        }
-       if ((vp->v_lflag & (VL_MARKTERM | VL_TERMINATE | VL_DEAD | VL_NEEDINACTIVE)) == VL_NEEDINACTIVE) {
+       if ((vp->v_lflag & (VL_DEAD | VL_NEEDINACTIVE)) == VL_NEEDINACTIVE) {
 
                vp->v_lflag &= ~VL_NEEDINACTIVE;
                vnode_unlock(vp);
 
                vp->v_lflag &= ~VL_NEEDINACTIVE;
                vnode_unlock(vp);
@@ -3914,7 +3993,7 @@ vnode_drain(vnode_t vp)
 {
        
        if (vp->v_lflag & VL_DRAIN) {
 {
        
        if (vp->v_lflag & VL_DRAIN) {
-               panic("vnode_drain: recursuve drain");
+               panic("vnode_drain: recursive drain");
                return(ENOENT);
        }
        vp->v_lflag |= VL_DRAIN;
                return(ENOENT);
        }
        vp->v_lflag |= VL_DRAIN;
@@ -3922,13 +4001,16 @@ vnode_drain(vnode_t vp)
 
        while (vp->v_iocount > 1)
                msleep(&vp->v_iocount, &vp->v_lock, PVFS, "vnode_drain", NULL);
 
        while (vp->v_iocount > 1)
                msleep(&vp->v_iocount, &vp->v_lock, PVFS, "vnode_drain", NULL);
+
+       vp->v_lflag &= ~VL_DRAIN;
+
        return(0);
 }
 
 
 /*
  * if the number of recent references via vnode_getwithvid or vnode_getwithref
        return(0);
 }
 
 
 /*
  * if the number of recent references via vnode_getwithvid or vnode_getwithref
- * exceeds this threshhold, than 'UN-AGE' the vnode by removing it from
+ * exceeds this threshold, than 'UN-AGE' the vnode by removing it from
  * the LRU list if it's currently on it... once the iocount and usecount both drop
  * to 0, it will get put back on the end of the list, effectively making it younger
  * this allows us to keep actively referenced vnodes in the list without having
  * the LRU list if it's currently on it... once the iocount and usecount both drop
  * to 0, it will get put back on the end of the list, effectively making it younger
  * this allows us to keep actively referenced vnodes in the list without having
@@ -3937,12 +4019,13 @@ vnode_drain(vnode_t vp)
  */
 #define UNAGE_THRESHHOLD       25
 
  */
 #define UNAGE_THRESHHOLD       25
 
-static errno_t
+errno_t
 vnode_getiocount(vnode_t vp, unsigned int vid, int vflags)
 {
        int nodead = vflags & VNODE_NODEAD;
        int nosusp = vflags & VNODE_NOSUSPEND;
        int always = vflags & VNODE_ALWAYS;
 vnode_getiocount(vnode_t vp, unsigned int vid, int vflags)
 {
        int nodead = vflags & VNODE_NODEAD;
        int nosusp = vflags & VNODE_NOSUSPEND;
        int always = vflags & VNODE_ALWAYS;
+       int beatdrain = vflags & VNODE_DRAINO;
 
        for (;;) {
                /*
 
        for (;;) {
                /*
@@ -3974,6 +4057,18 @@ vnode_getiocount(vnode_t vp, unsigned int vid, int vflags)
                
                if (always != 0) 
                        break;
                
                if (always != 0) 
                        break;
+
+               /*
+                * In some situations, we want to get an iocount
+                * even if the vnode is draining to prevent deadlock,
+                * e.g. if we're in the filesystem, potentially holding
+                * resources that could prevent other iocounts from
+                * being released.
+                */
+               if (beatdrain && (vp->v_lflag & VL_DRAIN)) {
+                       break;
+               }
+
                vnode_lock_convert(vp);
 
                if (vp->v_lflag & VL_TERMINATE) {
                vnode_lock_convert(vp);
 
                if (vp->v_lflag & VL_TERMINATE) {
@@ -3983,7 +4078,7 @@ vnode_getiocount(vnode_t vp, unsigned int vid, int vflags)
                } else
                        msleep(&vp->v_iocount, &vp->v_lock, PVFS, "vnode_getiocount", NULL);
        }
                } else
                        msleep(&vp->v_iocount, &vp->v_lock, PVFS, "vnode_getiocount", NULL);
        }
-       if (vid != vp->v_id) {
+       if (((vflags & VNODE_WITHID) != 0) && vid != vp->v_id) {
                return(ENOENT);
        }
        if (++vp->v_references >= UNAGE_THRESHHOLD) {
                return(ENOENT);
        }
        if (++vp->v_references >= UNAGE_THRESHHOLD) {
@@ -4087,7 +4182,6 @@ vnode_reclaim_internal(struct vnode * vp, int locked, int reuse, int flags)
        vp->v_socket = NULL;
 
        vp->v_lflag &= ~VL_TERMINATE;
        vp->v_socket = NULL;
 
        vp->v_lflag &= ~VL_TERMINATE;
-       vp->v_lflag &= ~VL_DRAIN;
        vp->v_owner = NULL;
 
        KNOTE(&vp->v_knotes, NOTE_REVOKE);
        vp->v_owner = NULL;
 
        KNOTE(&vp->v_knotes, NOTE_REVOKE);
@@ -4114,7 +4208,6 @@ vnode_reclaim_internal(struct vnode * vp, int locked, int reuse, int flags)
  * The following api creates a vnode and associates all the parameter specified in vnode_fsparam
  * structure and returns a vnode handle with a reference. device aliasing is handled here so checkalias
  * is obsoleted by this.
  * The following api creates a vnode and associates all the parameter specified in vnode_fsparam
  * structure and returns a vnode handle with a reference. device aliasing is handled here so checkalias
  * is obsoleted by this.
- *  vnode_create(int flavor, size_t size, void * param,  vnode_t  *vp)
  */
 int  
 vnode_create(uint32_t flavor, uint32_t size, void *data, vnode_t *vpp)
  */
 int  
 vnode_create(uint32_t flavor, uint32_t size, void *data, vnode_t *vpp)
@@ -4127,159 +4220,210 @@ vnode_create(uint32_t flavor, uint32_t size, void *data, vnode_t *vpp)
         struct  uthread *ut;
        struct componentname *cnp;
        struct vnode_fsparam *param = (struct vnode_fsparam *)data;
         struct  uthread *ut;
        struct componentname *cnp;
        struct vnode_fsparam *param = (struct vnode_fsparam *)data;
-       
-       if (flavor == VNCREATE_FLAVOR && (size == VCREATESIZE) && param) {
-               if ( (error = new_vnode(&vp)) ) {
-                       return(error);
-               } else {
-                       dvp = param->vnfs_dvp;
-                       cnp = param->vnfs_cnp;
+#if CONFIG_TRIGGERS
+       struct vnode_trigger_param *tinfo = NULL;
+#endif
+       if (param == NULL)
+               return (EINVAL);
 
 
-                       vp->v_op = param->vnfs_vops;
-                       vp->v_type = param->vnfs_vtype;
-                       vp->v_data = param->vnfs_fsnode;
+#if CONFIG_TRIGGERS
+       if ((flavor == VNCREATE_TRIGGER) && (size == VNCREATE_TRIGGER_SIZE)) {
+               tinfo = (struct vnode_trigger_param *)data;
+
+               /* Validate trigger vnode input */
+               if ((param->vnfs_vtype != VDIR) ||
+                   (tinfo->vnt_resolve_func == NULL) ||
+                   (tinfo->vnt_flags & ~VNT_VALID_MASK)) {
+                       return (EINVAL);
+               }
+               /* Fall through a normal create (params will be the same) */
+               flavor = VNCREATE_FLAVOR;
+               size = VCREATESIZE;
+       }
+#endif
+       if ((flavor != VNCREATE_FLAVOR) || (size != VCREATESIZE))
+               return (EINVAL);
+
+       if ( (error = new_vnode(&vp)) )
+               return(error);
 
 
-                       if (param->vnfs_markroot)
-                               vp->v_flag |= VROOT;
-                       if (param->vnfs_marksystem)
-                               vp->v_flag |= VSYSTEM;
-                       if (vp->v_type == VREG) {
-                               error = ubc_info_init_withsize(vp, param->vnfs_filesize);
-                               if (error) {
+       dvp = param->vnfs_dvp;
+       cnp = param->vnfs_cnp;
+
+       vp->v_op = param->vnfs_vops;
+       vp->v_type = param->vnfs_vtype;
+       vp->v_data = param->vnfs_fsnode;
+
+       if (param->vnfs_markroot)
+               vp->v_flag |= VROOT;
+       if (param->vnfs_marksystem)
+               vp->v_flag |= VSYSTEM;
+       if (vp->v_type == VREG) {
+               error = ubc_info_init_withsize(vp, param->vnfs_filesize);
+               if (error) {
 #ifdef JOE_DEBUG
 #ifdef JOE_DEBUG
-                                       record_vp(vp, 1);
+                       record_vp(vp, 1);
 #endif
 #endif
-                                       vp->v_mount = NULL;
-                                       vp->v_op = dead_vnodeop_p;
-                                       vp->v_tag = VT_NON;
-                                       vp->v_data = NULL;
-                                       vp->v_type = VBAD;
-                                       vp->v_lflag |= VL_DEAD;
-
-                                       vnode_put(vp);
-                                       return(error);
-                               }
-                       }
+                       vp->v_mount = NULL;
+                       vp->v_op = dead_vnodeop_p;
+                       vp->v_tag = VT_NON;
+                       vp->v_data = NULL;
+                       vp->v_type = VBAD;
+                       vp->v_lflag |= VL_DEAD;
+
+                       vnode_put(vp);
+                       return(error);
+               }
+       }
+#ifdef JOE_DEBUG
+       record_vp(vp, 1);
+#endif
+
+#if CONFIG_TRIGGERS
+       /*
+        * For trigger vnodes, attach trigger info to vnode
+        */
+       if ((vp->v_type == VDIR) && (tinfo != NULL)) {
+               /* 
+                * Note: has a side effect of incrementing trigger count on the
+                * mount if successful, which we would need to undo on a 
+                * subsequent failure.
+                */
+#ifdef JOE_DEBUG
+               record_vp(vp, -1);
+#endif
+               error = vnode_resolver_create(param->vnfs_mp, vp, tinfo, FALSE);
+               if (error) {
+                       printf("vnode_create: vnode_resolver_create() err %d\n", error);
+                       vp->v_mount = NULL;
+                       vp->v_op = dead_vnodeop_p;
+                       vp->v_tag = VT_NON;
+                       vp->v_data = NULL;
+                       vp->v_type = VBAD;
+                       vp->v_lflag |= VL_DEAD;
 #ifdef JOE_DEBUG
                        record_vp(vp, 1);
 #endif
 #ifdef JOE_DEBUG
                        record_vp(vp, 1);
 #endif
-                       if (vp->v_type == VCHR || vp->v_type == VBLK) {
-                
-                               vp->v_tag = VT_DEVFS;           /* callers will reset if needed (bdevvp) */
-
-                               if ( (nvp = checkalias(vp, param->vnfs_rdev)) ) {
-                                       /*
-                                        * if checkalias returns a vnode, it will be locked
-                                        *
-                                        * first get rid of the unneeded vnode we acquired
-                                        */
-                                       vp->v_data = NULL;
-                                       vp->v_op = spec_vnodeop_p;
-                                       vp->v_type = VBAD;
-                                       vp->v_lflag = VL_DEAD;
-                                       vp->v_data = NULL; 
-                                       vp->v_tag = VT_NON;
-                                       vnode_put(vp);
+                       vnode_put(vp);
+                       return (error);
+               }
+       }
+#endif
+       if (vp->v_type == VCHR || vp->v_type == VBLK) {
 
 
-                                       /*
-                                        * switch to aliased vnode and finish
-                                        * preparing it
-                                        */
-                                       vp = nvp;
-
-                                       vclean(vp, 0);
-                                       vp->v_op = param->vnfs_vops;
-                                       vp->v_type = param->vnfs_vtype;
-                                       vp->v_data = param->vnfs_fsnode;
-                                       vp->v_lflag = 0;
-                                       vp->v_mount = NULL;
-                                       insmntque(vp, param->vnfs_mp);
-                                       insert = 0;
-                                       vnode_unlock(vp);
-                               }
-                       }
+               vp->v_tag = VT_DEVFS;           /* callers will reset if needed (bdevvp) */
 
 
-                       if (vp->v_type == VFIFO) {
-                               struct fifoinfo *fip;
+               if ( (nvp = checkalias(vp, param->vnfs_rdev)) ) {
+                       /*
+                        * if checkalias returns a vnode, it will be locked
+                        *
+                        * first get rid of the unneeded vnode we acquired
+                        */
+                       vp->v_data = NULL;
+                       vp->v_op = spec_vnodeop_p;
+                       vp->v_type = VBAD;
+                       vp->v_lflag = VL_DEAD;
+                       vp->v_data = NULL; 
+                       vp->v_tag = VT_NON;
+                       vnode_put(vp);
 
 
-                               MALLOC(fip, struct fifoinfo *,
-                                       sizeof(*fip), M_TEMP, M_WAITOK);
-                               bzero(fip, sizeof(struct fifoinfo ));
-                               vp->v_fifoinfo = fip;
-                       }
-                       /* The file systems must pass the address of the location where
-                        * they store the vnode pointer. When we add the vnode into the mount
-                        * list and name cache they become discoverable. So the file system node
-                        * must have the connection to vnode setup by then
+                       /*
+                        * switch to aliased vnode and finish
+                        * preparing it
                         */
                         */
-                       *vpp = vp;
+                       vp = nvp;
 
 
-                       /* Add fs named reference. */
-                       if (param->vnfs_flags & VNFS_ADDFSREF) {
-                               vp->v_lflag |= VNAMED_FSHASH;
-                       }
-                       if (param->vnfs_mp) {
-                                       if (param->vnfs_mp->mnt_kern_flag & MNTK_LOCK_LOCAL)
-                                               vp->v_flag |= VLOCKLOCAL;
-                               if (insert) {
-                                       if ((vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb))
-                                               panic("insmntque: vp on the free list\n");
-                                       /*
-                                        * enter in mount vnode list
-                                        */
-                                       insmntque(vp, param->vnfs_mp);
-                               }
+                       vclean(vp, 0);
+                       vp->v_op = param->vnfs_vops;
+                       vp->v_type = param->vnfs_vtype;
+                       vp->v_data = param->vnfs_fsnode;
+                       vp->v_lflag = 0;
+                       vp->v_mount = NULL;
+                       insmntque(vp, param->vnfs_mp);
+                       insert = 0;
+                       vnode_unlock(vp);
+               }
+       }
+
+       if (vp->v_type == VFIFO) {
+               struct fifoinfo *fip;
+
+               MALLOC(fip, struct fifoinfo *,
+                       sizeof(*fip), M_TEMP, M_WAITOK);
+               bzero(fip, sizeof(struct fifoinfo ));
+               vp->v_fifoinfo = fip;
+       }
+       /* The file systems must pass the address of the location where
+        * they store the vnode pointer. When we add the vnode into the mount
+        * list and name cache they become discoverable. So the file system node
+        * must have the connection to vnode setup by then
+        */
+       *vpp = vp;
+
+       /* Add fs named reference. */
+       if (param->vnfs_flags & VNFS_ADDFSREF) {
+               vp->v_lflag |= VNAMED_FSHASH;
+       }
+       if (param->vnfs_mp) {
+                       if (param->vnfs_mp->mnt_kern_flag & MNTK_LOCK_LOCAL)
+                               vp->v_flag |= VLOCKLOCAL;
+               if (insert) {
+                       if ((vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb))
+                               panic("insmntque: vp on the free list\n");
+
+                       /*
+                        * enter in mount vnode list
+                        */
+                       insmntque(vp, param->vnfs_mp);
+               }
 #ifndef __LP64__
 #ifndef __LP64__
-                               if ((param->vnfs_mp->mnt_vtable->vfc_vfsflags & VFC_VFSTHREADSAFE) == 0) {
-                                       MALLOC_ZONE(vp->v_unsafefs, struct unsafe_fsnode *,
-                                                   sizeof(struct unsafe_fsnode), M_UNSAFEFS, M_WAITOK);
-                                       vp->v_unsafefs->fsnode_count = 0;
-                                       vp->v_unsafefs->fsnodeowner  = (void *)NULL;
-                                       lck_mtx_init(&vp->v_unsafefs->fsnodelock, vnode_lck_grp, vnode_lck_attr);
-                               }
+               if ((param->vnfs_mp->mnt_vtable->vfc_vfsflags & VFC_VFSTHREADSAFE) == 0) {
+                       MALLOC_ZONE(vp->v_unsafefs, struct unsafe_fsnode *,
+                                   sizeof(struct unsafe_fsnode), M_UNSAFEFS, M_WAITOK);
+                       vp->v_unsafefs->fsnode_count = 0;
+                       vp->v_unsafefs->fsnodeowner  = (void *)NULL;
+                       lck_mtx_init(&vp->v_unsafefs->fsnodelock, vnode_lck_grp, vnode_lck_attr);
+               }
 #endif /* __LP64__ */
 #endif /* __LP64__ */
-                       }
-                       if (dvp && vnode_ref(dvp) == 0) {
-                               vp->v_parent = dvp;
-                       }
-                       if (cnp) {
-                               if (dvp && ((param->vnfs_flags & (VNFS_NOCACHE | VNFS_CANTCACHE)) == 0)) {
-                                       /*
-                                        * enter into name cache
-                                        * we've got the info to enter it into the name cache now
-                                        * cache_enter_create will pick up an extra reference on
-                                        * the name entered into the string cache
-                                        */
-                                       vp->v_name = cache_enter_create(dvp, vp, cnp);
-                               } else
-                                       vp->v_name = vfs_addname(cnp->cn_nameptr, cnp->cn_namelen, cnp->cn_hash, 0);
+       }
+       if (dvp && vnode_ref(dvp) == 0) {
+               vp->v_parent = dvp;
+       }
+       if (cnp) {
+               if (dvp && ((param->vnfs_flags & (VNFS_NOCACHE | VNFS_CANTCACHE)) == 0)) {
+                       /*
+                        * enter into name cache
+                        * we've got the info to enter it into the name cache now
+                        * cache_enter_create will pick up an extra reference on
+                        * the name entered into the string cache
+                        */
+                       vp->v_name = cache_enter_create(dvp, vp, cnp);
+               } else
+                       vp->v_name = vfs_addname(cnp->cn_nameptr, cnp->cn_namelen, cnp->cn_hash, 0);
 
 
-                               if ((cnp->cn_flags & UNIONCREATED) == UNIONCREATED)
-                                       vp->v_flag |= VISUNION;
-                       }
-                       if ((param->vnfs_flags & VNFS_CANTCACHE) == 0) {
-                               /*
-                                * this vnode is being created as cacheable in the name cache
-                                * this allows us to re-enter it in the cache
-                                */
-                               vp->v_flag |= VNCACHEABLE;
-                       }
-                       ut = get_bsdthread_info(current_thread());
-               
-                       if ((current_proc()->p_lflag & P_LRAGE_VNODES) ||
-                           (ut->uu_flag & UT_RAGE_VNODES)) {
-                               /*
-                                * process has indicated that it wants any
-                                * vnodes created on its behalf to be rapidly
-                                * aged to reduce the impact on the cached set
-                                * of vnodes
-                                */
-                               vp->v_flag |= VRAGE;
-                       }
-                       return(0);
-               }
+               if ((cnp->cn_flags & UNIONCREATED) == UNIONCREATED)
+                       vp->v_flag |= VISUNION;
+       }
+       if ((param->vnfs_flags & VNFS_CANTCACHE) == 0) {
+               /*
+                * this vnode is being created as cacheable in the name cache
+                * this allows us to re-enter it in the cache
+                */
+               vp->v_flag |= VNCACHEABLE;
+       }
+       ut = get_bsdthread_info(current_thread());
+
+       if ((current_proc()->p_lflag & P_LRAGE_VNODES) ||
+           (ut->uu_flag & UT_RAGE_VNODES)) {
+               /*
+                * process has indicated that it wants any
+                * vnodes created on its behalf to be rapidly
+                * aged to reduce the impact on the cached set
+                * of vnodes
+                */
+               vp->v_flag |= VRAGE;
        }
        }
-       return (EINVAL);
+       return (0);
 }
 
 int
 }
 
 int
@@ -4309,13 +4453,14 @@ vnode_removefsref(vnode_t vp)
 
 
 int
 
 
 int
-vfs_iterate(__unused int flags, int (*callout)(mount_t, void *), void *arg)
+vfs_iterate(int flags, int (*callout)(mount_t, void *), void *arg)
 {
        mount_t mp;
        int ret = 0;
        fsid_t * fsid_list;
        int count, actualcount,  i;
        void * allocmem;
 {
        mount_t mp;
        int ret = 0;
        fsid_t * fsid_list;
        int count, actualcount,  i;
        void * allocmem;
+       int indx_start, indx_stop, indx_incr;
 
        count = mount_getvfscnt();
        count += 10;
 
        count = mount_getvfscnt();
        count += 10;
@@ -4325,7 +4470,21 @@ vfs_iterate(__unused int flags, int (*callout)(mount_t, void *), void *arg)
 
        actualcount = mount_fillfsids(fsid_list, count);
 
 
        actualcount = mount_fillfsids(fsid_list, count);
 
-       for (i=0; i< actualcount; i++) {
+       /*
+        * Establish the iteration direction
+        * VFS_ITERATE_TAIL_FIRST overrides default head first order (oldest first)
+        */
+       if (flags & VFS_ITERATE_TAIL_FIRST) {
+               indx_start = actualcount - 1;
+               indx_stop = -1;
+               indx_incr = -1;
+       } else /* Head first by default */ {
+               indx_start = 0;
+               indx_stop = actualcount;
+               indx_incr = 1;
+       }
+
+       for (i=indx_start; i != indx_stop; i += indx_incr) {
 
                /* obtain the mount point with iteration reference */
                mp = mount_list_lookupby_fsid(&fsid_list[i], 0, 1);
 
                /* obtain the mount point with iteration reference */
                mp = mount_list_lookupby_fsid(&fsid_list[i], 0, 1);
@@ -4567,7 +4726,8 @@ vnode_lookup(const char *path, int flags, vnode_t *vpp, vfs_context_t ctx)
                ndflags |= DOWHITEOUT;
 
        /* XXX AUDITVNPATH1 needed ? */
                ndflags |= DOWHITEOUT;
 
        /* XXX AUDITVNPATH1 needed ? */
-       NDINIT(&nd, LOOKUP, ndflags, UIO_SYSSPACE, CAST_USER_ADDR_T(path), ctx);
+       NDINIT(&nd, LOOKUP, OP_LOOKUP, ndflags, UIO_SYSSPACE,
+              CAST_USER_ADDR_T(path), ctx);
 
        if ((error = namei(&nd)))
                return (error);
 
        if ((error = namei(&nd)))
                return (error);
@@ -4603,7 +4763,8 @@ vnode_open(const char *path, int fmode, int cmode, int flags, vnode_t *vpp, vfs_
                ndflags |= DOWHITEOUT;
        
        /* XXX AUDITVNPATH1 needed ? */
                ndflags |= DOWHITEOUT;
        
        /* XXX AUDITVNPATH1 needed ? */
-       NDINIT(&nd, LOOKUP, ndflags, UIO_SYSSPACE, CAST_USER_ADDR_T(path), ctx);
+       NDINIT(&nd, LOOKUP, OP_OPEN, ndflags, UIO_SYSSPACE,
+              CAST_USER_ADDR_T(path), ctx);
 
        if ((error = vn_open(&nd, fmode, cmode)))
                *vpp = NULL;
 
        if ((error = vn_open(&nd, fmode, cmode)))
                *vpp = NULL;
@@ -4656,6 +4817,18 @@ vnode_setsize(vnode_t vp, off_t size, int ioflag, vfs_context_t ctx)
        return(vnode_setattr(vp, &va, ctx));
 }
 
        return(vnode_setattr(vp, &va, ctx));
 }
 
+static int
+vn_create_reg(vnode_t dvp, vnode_t *vpp, struct nameidata *ndp, struct vnode_attr *vap, uint32_t flags, int fmode, uint32_t *statusp, vfs_context_t ctx)
+{
+       /* Only use compound VNOP for compound operation */
+       if (vnode_compound_open_available(dvp) && ((flags & VN_CREATE_DOOPEN) != 0)) {
+               *vpp = NULLVP;
+               return VNOP_COMPOUND_OPEN(dvp, vpp, ndp, VNOP_COMPOUND_OPEN_DO_CREATE, fmode, statusp, vap, ctx);
+       } else {
+               return VNOP_CREATE(dvp, vpp, &ndp->ni_cnd, vap, ctx);
+       }
+}
+
 /*
  * Create a filesystem object of arbitrary type with arbitrary attributes in
  * the spevied directory with the specified name.
 /*
  * Create a filesystem object of arbitrary type with arbitrary attributes in
  * the spevied directory with the specified name.
@@ -4698,70 +4871,48 @@ vnode_setsize(vnode_t vp, off_t size, int ioflag, vfs_context_t ctx)
  *             in the code they originated.
  */
 errno_t
  *             in the code they originated.
  */
 errno_t
-vn_create(vnode_t dvp, vnode_t *vpp, struct componentname *cnp, struct vnode_attr *vap, int flags, vfs_context_t ctx)
+vn_create(vnode_t dvp, vnode_t *vpp, struct nameidata *ndp, struct vnode_attr *vap, uint32_t flags, int fmode, uint32_t *statusp, vfs_context_t ctx)
 {
 {
-       kauth_acl_t oacl, nacl;
-       int initial_acl;
-       errno_t error;
+       errno_t error, old_error;
        vnode_t vp = (vnode_t)0;
        vnode_t vp = (vnode_t)0;
+       boolean_t batched;
+       struct componentname *cnp;
+       uint32_t defaulted;
 
 
+       cnp = &ndp->ni_cnd;
        error = 0;
        error = 0;
-       oacl = nacl = NULL;
-       initial_acl = 0;
+       batched = namei_compound_available(dvp, ndp) ? TRUE : FALSE;
 
        KAUTH_DEBUG("%p    CREATE - '%s'", dvp, cnp->cn_nameptr);
 
 
        KAUTH_DEBUG("%p    CREATE - '%s'", dvp, cnp->cn_nameptr);
 
+       if (flags & VN_CREATE_NOINHERIT) 
+               vap->va_vaflags |= VA_NOINHERIT;
+       if (flags & VN_CREATE_NOAUTH) 
+               vap->va_vaflags |= VA_NOAUTH;
        /*
        /*
-        * Handle ACL inheritance.
+        * Handle ACL inheritance, initialize vap.
         */
         */
-       if (!(flags & VN_CREATE_NOINHERIT) && vfs_extendedsecurity(dvp->v_mount)) {
-               /* save the original filesec */
-               if (VATTR_IS_ACTIVE(vap, va_acl)) {
-                       initial_acl = 1;
-                       oacl = vap->va_acl;
-               }
-
-               vap->va_acl = NULL;
-               if ((error = kauth_acl_inherit(dvp,
-                        oacl,
-                        &nacl,
-                        vap->va_type == VDIR,
-                        ctx)) != 0) {
-                       KAUTH_DEBUG("%p    CREATE - error %d processing inheritance", dvp, error);
-                       return(error);
-               }
+       error = vn_attribute_prepare(dvp, vap, &defaulted, ctx);
+       if (error) {
+               return error;
+       }
 
 
-               /*
-                * If the generated ACL is NULL, then we can save ourselves some effort
-                * by clearing the active bit.
-                */
-               if (nacl == NULL) {
-                       VATTR_CLEAR_ACTIVE(vap, va_acl);
-               } else {
-                       VATTR_SET(vap, va_acl, nacl);
-               }
+       if (vap->va_type != VREG && (fmode != 0 || (flags & VN_CREATE_DOOPEN) || statusp)) {
+               panic("Open parameters, but not a regular file.");
        }
        }
-       
-       /*
-        * Check and default new attributes.
-        * This will set va_uid, va_gid, va_mode and va_create_time at least, if the caller
-        * hasn't supplied them.
-        */
-       if ((error = vnode_authattr_new(dvp, vap, flags & VN_CREATE_NOAUTH, ctx)) != 0) {
-               KAUTH_DEBUG("%p    CREATE - error %d handing/defaulting attributes", dvp, error);
-               goto out;
+       if ((fmode != 0) && ((flags & VN_CREATE_DOOPEN) == 0)) {
+               panic("Mode for open, but not trying to open...");
        }
 
        }
 
-               
        /*
         * Create the requested node.
         */
        switch(vap->va_type) {
        case VREG:
        /*
         * Create the requested node.
         */
        switch(vap->va_type) {
        case VREG:
-               error = VNOP_CREATE(dvp, vpp, cnp, vap, ctx);
+               error = vn_create_reg(dvp, vpp, ndp, vap, flags, fmode, statusp, ctx);
                break;
        case VDIR:
                break;
        case VDIR:
-               error = VNOP_MKDIR(dvp, vpp, cnp, vap, ctx);
+               error = vn_mkdir(dvp, vpp, ndp, vap, ctx);
                break;
        case VSOCK:
        case VFIFO:
                break;
        case VSOCK:
        case VFIFO:
@@ -4778,6 +4929,8 @@ vn_create(vnode_t dvp, vnode_t *vpp, struct componentname *cnp, struct vnode_att
        }
 
        vp = *vpp;
        }
 
        vp = *vpp;
+       old_error = error;
+
 #if CONFIG_MACF
        if (!(flags & VN_CREATE_NOLABEL)) {
                error = vnode_label(vnode_mount(vp), dvp, vp, cnp, VNODE_LABEL_CREATE, ctx);
 #if CONFIG_MACF
        if (!(flags & VN_CREATE_NOLABEL)) {
                error = vnode_label(vnode_mount(vp), dvp, vp, cnp, VNODE_LABEL_CREATE, ctx);
@@ -4797,24 +4950,22 @@ vn_create(vnode_t dvp, vnode_t *vpp, struct componentname *cnp, struct vnode_att
 #if CONFIG_MACF
 error:
 #endif
 #if CONFIG_MACF
 error:
 #endif
-       if ((error != 0 ) && (vp != (vnode_t)0)) {
-               *vpp = (vnode_t) 0;
-               vnode_put(vp);
+       if ((error != 0) && (vp != (vnode_t)0)) {
+
+               /* If we've done a compound open, close */
+               if (batched && (old_error == 0) && (vap->va_type == VREG)) {
+                       VNOP_CLOSE(vp, fmode, ctx);
+               }
+
+               /* Need to provide notifications if a create succeeded */
+               if (!batched) {
+                       *vpp = (vnode_t) 0;
+                       vnode_put(vp);
+               }
        }
 
 out:
        }
 
 out:
-       /*
-        * If the caller supplied a filesec in vap, it has been replaced
-        * now by the post-inheritance copy.  We need to put the original back
-        * and free the inherited product.
-        */
-       if (initial_acl) {
-               VATTR_SET(vap, va_acl, oacl);
-       } else {
-               VATTR_CLEAR_ACTIVE(vap, va_acl);
-       }
-       if (nacl != NULL)
-               kauth_acl_free(nacl);
+       vn_attribute_cleanup(vap, defaulted);
 
        return(error);
 }
 
        return(error);
 }
@@ -4845,6 +4996,433 @@ vnode_authorize_init(void)
        vnode_scope = kauth_register_scope(KAUTH_SCOPE_VNODE, vnode_authorize_callback, NULL);
 }
 
        vnode_scope = kauth_register_scope(KAUTH_SCOPE_VNODE, vnode_authorize_callback, NULL);
 }
 
+#define VATTR_PREPARE_DEFAULTED_UID            0x1
+#define VATTR_PREPARE_DEFAULTED_GID            0x2
+#define VATTR_PREPARE_DEFAULTED_MODE           0x4
+
+int
+vn_attribute_prepare(vnode_t dvp, struct vnode_attr *vap, uint32_t *defaulted_fieldsp, vfs_context_t ctx)
+{
+       kauth_acl_t nacl = NULL, oacl = NULL;
+       int error;
+
+       /*
+        * Handle ACL inheritance.
+        */
+       if (!(vap->va_vaflags & VA_NOINHERIT) && vfs_extendedsecurity(dvp->v_mount)) {
+               /* save the original filesec */
+               if (VATTR_IS_ACTIVE(vap, va_acl)) {
+                       oacl = vap->va_acl;
+               }
+
+               vap->va_acl = NULL;
+               if ((error = kauth_acl_inherit(dvp,
+                        oacl,
+                        &nacl,
+                        vap->va_type == VDIR,
+                        ctx)) != 0) {
+                       KAUTH_DEBUG("%p    CREATE - error %d processing inheritance", dvp, error);
+                       return(error);
+               }
+
+               /*
+                * If the generated ACL is NULL, then we can save ourselves some effort
+                * by clearing the active bit.
+                */
+               if (nacl == NULL) {
+                       VATTR_CLEAR_ACTIVE(vap, va_acl);
+               } else {
+                       vap->va_base_acl = oacl;
+                       VATTR_SET(vap, va_acl, nacl);
+               }
+       }
+       
+       error = vnode_authattr_new_internal(dvp, vap, (vap->va_vaflags & VA_NOAUTH), defaulted_fieldsp, ctx);
+       if (error) {
+               vn_attribute_cleanup(vap, *defaulted_fieldsp);
+       } 
+
+       return error;
+}
+
+void
+vn_attribute_cleanup(struct vnode_attr *vap, uint32_t defaulted_fields)
+{
+       /*
+        * If the caller supplied a filesec in vap, it has been replaced
+        * now by the post-inheritance copy.  We need to put the original back
+        * and free the inherited product.
+        */
+       kauth_acl_t nacl, oacl;
+
+       if (VATTR_IS_ACTIVE(vap, va_acl)) {
+               nacl = vap->va_acl;
+               oacl = vap->va_base_acl;
+
+               if (oacl)  {
+                       VATTR_SET(vap, va_acl, oacl);
+                       vap->va_base_acl = NULL;
+               } else {
+                       VATTR_CLEAR_ACTIVE(vap, va_acl);
+               }
+
+               if (nacl != NULL) {
+                       kauth_acl_free(nacl);
+               }
+       }
+
+       if ((defaulted_fields & VATTR_PREPARE_DEFAULTED_MODE) != 0) {
+               VATTR_CLEAR_ACTIVE(vap, va_mode);
+       }
+       if ((defaulted_fields & VATTR_PREPARE_DEFAULTED_GID) != 0) {
+               VATTR_CLEAR_ACTIVE(vap, va_gid);
+       }
+       if ((defaulted_fields & VATTR_PREPARE_DEFAULTED_UID) != 0) {
+               VATTR_CLEAR_ACTIVE(vap, va_uid);
+       }
+
+       return;
+}
+
+int
+vn_authorize_unlink(vnode_t dvp, vnode_t vp, struct componentname *cnp, vfs_context_t ctx, __unused void *reserved)
+{
+       int error = 0;
+
+       /*
+        * Normally, unlinking of directories is not supported. 
+        * However, some file systems may have limited support.
+        */
+       if ((vp->v_type == VDIR) &&
+                       !(vp->v_mount->mnt_vtable->vfc_vfsflags & VFC_VFSDIRLINKS)) {
+               return (EPERM); /* POSIX */
+       }
+
+       /* authorize the delete operation */
+#if CONFIG_MACF
+       if (!error)
+               error = mac_vnode_check_unlink(ctx, dvp, vp, cnp);
+#endif /* MAC */
+       if (!error)
+               error = vnode_authorize(vp, dvp, KAUTH_VNODE_DELETE, ctx);
+
+       return error;
+}
+
+int
+vn_authorize_open_existing(vnode_t vp, struct componentname *cnp, int fmode, vfs_context_t ctx, void *reserved)
+{
+       /* Open of existing case */
+       kauth_action_t action;
+       int error = 0;
+
+       if (cnp->cn_ndp == NULL) {
+               panic("NULL ndp");
+       }
+       if (reserved != NULL) {
+               panic("reserved not NULL.");
+       }
+
+#if CONFIG_MACF
+       /* XXX may do duplicate work here, but ignore that for now (idempotent) */
+       if (vfs_flags(vnode_mount(vp)) & MNT_MULTILABEL) {
+               error = vnode_label(vnode_mount(vp), NULL, vp, NULL, 0, ctx);
+               if (error)
+                       return (error);
+       }
+#endif
+
+       if ( (fmode & O_DIRECTORY) && vp->v_type != VDIR ) {
+               return (ENOTDIR);
+       }
+
+       if (vp->v_type == VSOCK && vp->v_tag != VT_FDESC) {
+               return (EOPNOTSUPP);    /* Operation not supported on socket */
+       }
+
+       if (vp->v_type == VLNK && (fmode & O_NOFOLLOW) != 0) {
+               return (ELOOP);         /* O_NOFOLLOW was specified and the target is a symbolic link */
+       }
+
+       /* disallow write operations on directories */
+       if (vnode_isdir(vp) && (fmode & (FWRITE | O_TRUNC))) {
+               return (EISDIR);
+       }
+
+       if ((cnp->cn_ndp->ni_flag & NAMEI_TRAILINGSLASH)) {
+               if (vp->v_type != VDIR) {
+                       return (ENOTDIR);
+               }
+       }
+
+#if CONFIG_MACF
+       /* If a file being opened is a shadow file containing 
+        * namedstream data, ignore the macf checks because it 
+        * is a kernel internal file and access should always 
+        * be allowed.
+        */
+       if (!(vnode_isshadow(vp) && vnode_isnamedstream(vp))) {
+               error = mac_vnode_check_open(ctx, vp, fmode);
+               if (error) {
+                       return (error);
+               }
+       }
+#endif
+
+       /* compute action to be authorized */
+       action = 0;
+       if (fmode & FREAD) {
+               action |= KAUTH_VNODE_READ_DATA;
+       }
+       if (fmode & (FWRITE | O_TRUNC)) {
+               /*
+                * If we are writing, appending, and not truncating,
+                * indicate that we are appending so that if the
+                * UF_APPEND or SF_APPEND bits are set, we do not deny
+                * the open.
+                */
+               if ((fmode & O_APPEND) && !(fmode & O_TRUNC)) {
+                       action |= KAUTH_VNODE_APPEND_DATA;
+               } else {
+                       action |= KAUTH_VNODE_WRITE_DATA;
+               }
+       }
+       return (vnode_authorize(vp, NULL, action, ctx));
+}
+
+int
+vn_authorize_create(vnode_t dvp, struct componentname *cnp, struct vnode_attr *vap, vfs_context_t ctx, void *reserved)
+{
+       /* Creation case */
+       int error;
+
+       if (cnp->cn_ndp == NULL) {
+               panic("NULL cn_ndp");
+       }
+       if (reserved != NULL) {
+               panic("reserved not NULL.");
+       }
+
+       /* Only validate path for creation if we didn't do a complete lookup */
+       if (cnp->cn_ndp->ni_flag & NAMEI_UNFINISHED) {
+               error = lookup_validate_creation_path(cnp->cn_ndp);
+               if (error)
+                       return (error);
+       }
+
+#if CONFIG_MACF
+       error = mac_vnode_check_create(ctx, dvp, cnp, vap);
+       if (error)
+               return (error);
+#endif /* CONFIG_MACF */
+
+       return (vnode_authorize(dvp, NULL, KAUTH_VNODE_ADD_FILE, ctx));
+}
+
+int 
+vn_authorize_rename(struct vnode *fdvp,  struct vnode *fvp,  struct componentname *fcnp,
+             struct vnode *tdvp,  struct vnode *tvp,  struct componentname *tcnp,
+             vfs_context_t ctx, void *reserved)
+{
+       int error = 0;
+       int moving = 0;
+
+       if (reserved != NULL) {
+               panic("Passed something other than NULL as reserved field!");
+       }
+
+       /*
+        * Avoid renaming "." and "..".
+        *
+        * XXX No need to check for this in the FS.  We should always have the leaves
+        * in VFS in this case.
+        */
+       if (fvp->v_type == VDIR &&
+           ((fdvp == fvp) ||
+            (fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
+            ((fcnp->cn_flags | tcnp->cn_flags) & ISDOTDOT)) ) {
+               error = EINVAL;
+               goto out;
+       }
+
+       if (tvp == NULLVP && vnode_compound_rename_available(tdvp)) {
+               error = lookup_validate_creation_path(tcnp->cn_ndp);
+               if (error) 
+                       goto out;
+       }
+
+       /***** <MACF> *****/
+#if CONFIG_MACF
+       error = mac_vnode_check_rename_from(ctx, fdvp, fvp, fcnp);
+       if (error)
+               goto out;
+#endif
+
+#if CONFIG_MACF
+       error = mac_vnode_check_rename_to(ctx,
+                       tdvp, tvp, fdvp == tdvp, tcnp);
+       if (error)
+               goto out;
+#endif
+       /***** </MACF> *****/
+
+       /***** <MiscChecks> *****/
+       if (tvp != NULL) {
+               if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
+                       error = ENOTDIR;
+                       goto out;
+               } else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
+                       error = EISDIR;
+                       goto out;
+               }
+       }
+
+       if (fvp == tdvp) {
+               error = EINVAL;
+               goto out;
+       }
+
+       /*
+        * The following edge case is caught here:
+        * (to cannot be a descendent of from)
+        *
+        *       o fdvp
+        *      /
+        *     /
+        *    o fvp
+        *     \
+        *      \
+        *       o tdvp
+        *      /
+        *     /
+        *    o tvp
+        */
+       if (tdvp->v_parent == fvp) {
+               error = EINVAL;
+               goto out;
+       }
+       /***** </MiscChecks> *****/
+
+       /***** <Kauth> *****/
+
+       error = 0;
+       if ((tvp != NULL) && vnode_isdir(tvp)) {
+               if (tvp != fdvp)
+                       moving = 1;
+       } else if (tdvp != fdvp) {
+               moving = 1;
+       }
+
+
+       /*
+        * must have delete rights to remove the old name even in
+        * the simple case of fdvp == tdvp.
+        *
+        * If fvp is a directory, and we are changing it's parent,
+        * then we also need rights to rewrite its ".." entry as well.
+        */
+       if (vnode_isdir(fvp)) {
+               if ((error = vnode_authorize(fvp, fdvp, KAUTH_VNODE_DELETE | KAUTH_VNODE_ADD_SUBDIRECTORY, ctx)) != 0)
+                       goto out;
+       } else {
+               if ((error = vnode_authorize(fvp, fdvp, KAUTH_VNODE_DELETE, ctx)) != 0)
+                       goto out;
+       }
+       if (moving) {
+               /* moving into tdvp or tvp, must have rights to add */
+               if ((error = vnode_authorize(((tvp != NULL) && vnode_isdir(tvp)) ? tvp : tdvp,
+                                               NULL, 
+                                               vnode_isdir(fvp) ? KAUTH_VNODE_ADD_SUBDIRECTORY : KAUTH_VNODE_ADD_FILE,
+                                               ctx)) != 0) {
+                       goto out;
+               }
+       } else {
+               /* node staying in same directory, must be allowed to add new name */
+               if ((error = vnode_authorize(fdvp, NULL,
+                                               vnode_isdir(fvp) ? KAUTH_VNODE_ADD_SUBDIRECTORY : KAUTH_VNODE_ADD_FILE, ctx)) != 0)
+                       goto out;
+       }
+       /* overwriting tvp */
+       if ((tvp != NULL) && !vnode_isdir(tvp) &&
+                       ((error = vnode_authorize(tvp, tdvp, KAUTH_VNODE_DELETE, ctx)) != 0)) {
+               goto out;
+       }
+
+       /***** </Kauth> *****/
+
+       /* XXX more checks? */
+out:
+       return error;
+}
+
+int
+vn_authorize_mkdir(vnode_t dvp, struct componentname *cnp, struct vnode_attr *vap, vfs_context_t ctx, void *reserved)
+{
+       int error;
+
+       if (reserved != NULL) {
+               panic("reserved not NULL in vn_authorize_mkdir()");     
+       }
+
+       /* XXX A hack for now, to make shadow files work */
+       if (cnp->cn_ndp == NULL) {
+               return 0;
+       }
+
+       if (vnode_compound_mkdir_available(dvp)) {
+               error = lookup_validate_creation_path(cnp->cn_ndp);
+               if (error)
+                       goto out;
+       }
+
+#if CONFIG_MACF
+       error = mac_vnode_check_create(ctx,
+           dvp, cnp, vap);
+       if (error)
+               goto out;
+#endif
+
+       /* authorize addition of a directory to the parent */
+       if ((error = vnode_authorize(dvp, NULL, KAUTH_VNODE_ADD_SUBDIRECTORY, ctx)) != 0)
+               goto out;
+       
+out:
+       return error;
+}
+
+int
+vn_authorize_rmdir(vnode_t dvp, vnode_t vp, struct componentname *cnp, vfs_context_t ctx, void *reserved)
+{
+       int error;
+
+       if (reserved != NULL) {
+               panic("Non-NULL reserved argument to vn_authorize_rmdir()");
+       }
+
+       if (vp->v_type != VDIR) {
+               /*
+                * rmdir only deals with directories
+                */
+               return ENOTDIR;
+       } 
+       
+       if (dvp == vp) {
+               /*
+                * No rmdir "." please.
+                */
+               return EINVAL;
+       } 
+       
+#if CONFIG_MACF
+       error = mac_vnode_check_unlink(ctx, dvp,
+                       vp, cnp);
+       if (error)
+               return error;
+#endif
+
+       return vnode_authorize(vp, dvp, KAUTH_VNODE_DELETE, ctx);
+}
+
 /*
  * Authorize an operation on a vnode.
  *
 /*
  * Authorize an operation on a vnode.
  *
@@ -5254,8 +5832,11 @@ vnode_authorize_posix(vauth_ctx vcp, int action, int on_dir)
        } else {
                error = vauth_file_ingroup(vcp, &ismember, (!group_ok ? EACCES : 0));
        }
        } else {
                error = vauth_file_ingroup(vcp, &ismember, (!group_ok ? EACCES : 0));
        }
-       if (error)
-               goto out;
+       if (error) {
+               if (!group_ok)
+                       ismember = 1;
+               error = 0;
+       }
        if (ismember) {
                _SETWHERE("group");
                if (!group_ok)
        if (ismember) {
                _SETWHERE("group");
                if (!group_ok)
@@ -5324,8 +5905,6 @@ vnode_authorize_delete(vauth_ctx vcp, boolean_t cached_delete_child)
        /* check the ACL on the directory */
        delete_child_denied = 0;
        if (!cached_delete_child && VATTR_IS_NOT(dvap, va_acl, NULL)) {
        /* check the ACL on the directory */
        delete_child_denied = 0;
        if (!cached_delete_child && VATTR_IS_NOT(dvap, va_acl, NULL)) {
-               errno_t posix_error;
-
                eval.ae_requested = KAUTH_VNODE_DELETE_CHILD;
                eval.ae_acl = &dvap->va_acl->acl_ace[0];
                eval.ae_count = dvap->va_acl->acl_entrycount;
                eval.ae_requested = KAUTH_VNODE_DELETE_CHILD;
                eval.ae_acl = &dvap->va_acl->acl_ace[0];
                eval.ae_count = dvap->va_acl->acl_entrycount;
@@ -5338,9 +5917,11 @@ vnode_authorize_delete(vauth_ctx vcp, boolean_t cached_delete_child)
                 * have the ACL evaluation answer.  Previously, we would
                 * always deny the operation at this point.
                 */
                 * have the ACL evaluation answer.  Previously, we would
                 * always deny the operation at this point.
                 */
-               if ((posix_error = vauth_dir_ingroup(vcp, &ismember, ENOENT)) != 0 && posix_error != ENOENT)
-                       return(posix_error);
-               if (ismember)
+               if ((error = vauth_dir_ingroup(vcp, &ismember, ENOENT)) != 0 && error != ENOENT)
+                       return(error);
+               if (error == ENOENT)
+                       eval.ae_options |= KAUTH_AEVAL_IN_GROUP_UNKNOWN;
+               else if (ismember)
                        eval.ae_options |= KAUTH_AEVAL_IN_GROUP;
                eval.ae_exp_gall = KAUTH_VNODE_GENERIC_ALL_BITS;
                eval.ae_exp_gread = KAUTH_VNODE_GENERIC_READ_BITS;
                        eval.ae_options |= KAUTH_AEVAL_IN_GROUP;
                eval.ae_exp_gall = KAUTH_VNODE_GENERIC_ALL_BITS;
                eval.ae_exp_gread = KAUTH_VNODE_GENERIC_READ_BITS;
@@ -5361,18 +5942,11 @@ vnode_authorize_delete(vauth_ctx vcp, boolean_t cached_delete_child)
                case KAUTH_RESULT_DENY:
                        delete_child_denied = 1;
                        break;
                case KAUTH_RESULT_DENY:
                        delete_child_denied = 1;
                        break;
-               case KAUTH_RESULT_ALLOW:
-                       KAUTH_DEBUG("%p    ALLOWED - granted by directory ACL", vcp->vp);
-                       return(0);
+                       /* FALLSTHROUGH */
+                case KAUTH_RESULT_ALLOW:
+                        KAUTH_DEBUG("%p    ALLOWED - granted by directory ACL", vcp->vp);
+                        return(0);
                case KAUTH_RESULT_DEFER:
                case KAUTH_RESULT_DEFER:
-                       /*
-                        * If we don't have a POSIX answer of "yes", and we
-                        * can't get an ACL answer, then we deny it now.
-                        */
-                       if (posix_error == ENOENT) {
-                               delete_child_denied = 1;
-                               break;
-                       }
                default:
                        /* Effectively the same as !delete_child_denied */
                        KAUTH_DEBUG("%p    DEFERRED - directory ACL", vcp->vp);
                default:
                        /* Effectively the same as !delete_child_denied */
                        KAUTH_DEBUG("%p    DEFERRED - directory ACL", vcp->vp);
@@ -5383,8 +5957,6 @@ vnode_authorize_delete(vauth_ctx vcp, boolean_t cached_delete_child)
        /* check the ACL on the node */
        delete_denied = 0;
        if (VATTR_IS_NOT(vap, va_acl, NULL)) {
        /* check the ACL on the node */
        delete_denied = 0;
        if (VATTR_IS_NOT(vap, va_acl, NULL)) {
-               errno_t posix_error;
-
                eval.ae_requested = KAUTH_VNODE_DELETE;
                eval.ae_acl = &vap->va_acl->acl_ace[0];
                eval.ae_count = vap->va_acl->acl_entrycount;
                eval.ae_requested = KAUTH_VNODE_DELETE;
                eval.ae_acl = &vap->va_acl->acl_ace[0];
                eval.ae_count = vap->va_acl->acl_entrycount;
@@ -5397,9 +5969,11 @@ vnode_authorize_delete(vauth_ctx vcp, boolean_t cached_delete_child)
                 * have the ACL evaluation answer.  Previously, we would
                 * always deny the operation at this point.
                 */
                 * have the ACL evaluation answer.  Previously, we would
                 * always deny the operation at this point.
                 */
-               if ((posix_error = vauth_file_ingroup(vcp, &ismember, ENOENT)) != 0 && posix_error != ENOENT)
-                       return(posix_error);
-               if (ismember)
+               if ((error = vauth_file_ingroup(vcp, &ismember, ENOENT)) != 0 && error != ENOENT)
+                       return(error);
+               if (error == ENOENT)
+                       eval.ae_options |= KAUTH_AEVAL_IN_GROUP_UNKNOWN;
+               else if (ismember)
                        eval.ae_options |= KAUTH_AEVAL_IN_GROUP;
                eval.ae_exp_gall = KAUTH_VNODE_GENERIC_ALL_BITS;
                eval.ae_exp_gread = KAUTH_VNODE_GENERIC_READ_BITS;
                        eval.ae_options |= KAUTH_AEVAL_IN_GROUP;
                eval.ae_exp_gall = KAUTH_VNODE_GENERIC_ALL_BITS;
                eval.ae_exp_gread = KAUTH_VNODE_GENERIC_READ_BITS;
@@ -5419,13 +5993,6 @@ vnode_authorize_delete(vauth_ctx vcp, boolean_t cached_delete_child)
                        KAUTH_DEBUG("%p    ALLOWED - granted by file ACL", vcp->vp);
                        return(0);
                case KAUTH_RESULT_DEFER:
                        KAUTH_DEBUG("%p    ALLOWED - granted by file ACL", vcp->vp);
                        return(0);
                case KAUTH_RESULT_DEFER:
-                       /*
-                        * If we don't have a POSIX answer of "yes", and we
-                        * can't get an ACL answer, then we deny it now.
-                        */
-                       if (posix_error == ENOENT) {
-                               delete_denied = 1;
-                       }
                default:
                        /* Effectively the same as !delete_child_denied */
                        KAUTH_DEBUG("%p    DEFERRED%s - by file ACL", vcp->vp, delete_denied ? "(DENY)" : "");
                default:
                        /* Effectively the same as !delete_child_denied */
                        KAUTH_DEBUG("%p    DEFERRED%s - by file ACL", vcp->vp, delete_denied ? "(DENY)" : "");
@@ -5447,13 +6014,13 @@ vnode_authorize_delete(vauth_ctx vcp, boolean_t cached_delete_child)
         */
        if (!cached_delete_child && (dvap->va_mode & S_ISTXT) && !vauth_file_owner(vcp) && !vauth_dir_owner(vcp)) {
                KAUTH_DEBUG("%p    DENIED - sticky bit rules (user %d  file %d  dir %d)",
         */
        if (!cached_delete_child && (dvap->va_mode & S_ISTXT) && !vauth_file_owner(vcp) && !vauth_dir_owner(vcp)) {
                KAUTH_DEBUG("%p    DENIED - sticky bit rules (user %d  file %d  dir %d)",
-                   vcp->vp, cred->cr_uid, vap->va_uid, dvap->va_uid);
+                   vcp->vp, cred->cr_posix.cr_uid, vap->va_uid, dvap->va_uid);
                return(EACCES);
        }
 
        /* check the directory */
        if (!cached_delete_child && (error = vnode_authorize_posix(vcp, VWRITE, 1 /* on_dir */)) != 0) {
                return(EACCES);
        }
 
        /* check the directory */
        if (!cached_delete_child && (error = vnode_authorize_posix(vcp, VWRITE, 1 /* on_dir */)) != 0) {
-               KAUTH_DEBUG("%p    ALLOWED - granted by posix permisssions", vcp->vp);
+               KAUTH_DEBUG("%p    DENIED - denied by posix permisssions", vcp->vp);
                return(error);
        }
 
                return(error);
        }
 
@@ -5501,8 +6068,6 @@ vnode_authorize_simple(vauth_ctx vcp, kauth_ace_rights_t acl_rights, kauth_ace_r
 
        /* if we have an ACL, evaluate it */
        if (VATTR_IS_NOT(vap, va_acl, NULL)) {
 
        /* if we have an ACL, evaluate it */
        if (VATTR_IS_NOT(vap, va_acl, NULL)) {
-               errno_t posix_error;
-
                eval.ae_requested = acl_rights;
                eval.ae_acl = &vap->va_acl->acl_ace[0];
                eval.ae_count = vap->va_acl->acl_entrycount;
                eval.ae_requested = acl_rights;
                eval.ae_acl = &vap->va_acl->acl_ace[0];
                eval.ae_count = vap->va_acl->acl_entrycount;
@@ -5515,9 +6080,11 @@ vnode_authorize_simple(vauth_ctx vcp, kauth_ace_rights_t acl_rights, kauth_ace_r
                 * have the ACL evaluation answer.  Previously, we would
                 * always deny the operation at this point.
                 */
                 * have the ACL evaluation answer.  Previously, we would
                 * always deny the operation at this point.
                 */
-               if ((posix_error = vauth_file_ingroup(vcp, &ismember, ENOENT)) != 0 && posix_error != ENOENT)
-                       return(posix_error);
-               if (ismember)
+               if ((error = vauth_file_ingroup(vcp, &ismember, ENOENT)) != 0 && error != ENOENT)
+                       return(error);
+               if (error == ENOENT)
+                       eval.ae_options |= KAUTH_AEVAL_IN_GROUP_UNKNOWN;
+               else if (ismember)
                        eval.ae_options |= KAUTH_AEVAL_IN_GROUP;
                eval.ae_exp_gall = KAUTH_VNODE_GENERIC_ALL_BITS;
                eval.ae_exp_gread = KAUTH_VNODE_GENERIC_READ_BITS;
                        eval.ae_options |= KAUTH_AEVAL_IN_GROUP;
                eval.ae_exp_gall = KAUTH_VNODE_GENERIC_ALL_BITS;
                eval.ae_exp_gread = KAUTH_VNODE_GENERIC_READ_BITS;
@@ -5537,14 +6104,6 @@ vnode_authorize_simple(vauth_ctx vcp, kauth_ace_rights_t acl_rights, kauth_ace_r
                        KAUTH_DEBUG("%p    ALLOWED - all rights granted by ACL", vcp->vp);
                        return(0);
                case KAUTH_RESULT_DEFER:
                        KAUTH_DEBUG("%p    ALLOWED - all rights granted by ACL", vcp->vp);
                        return(0);
                case KAUTH_RESULT_DEFER:
-                       /*
-                        * If we don't have a POSIX answer of "yes", and we
-                        * can't get an ACL answer, then we deny it now.
-                        */
-                       if (posix_error == ENOENT) {
-                               KAUTH_DEBUG("%p    DENIED(DEFERRED) - by ACL", vcp->vp);
-                               return(EACCES);         /* deny, deny, counter-allege */
-                       }
                default:
                        /* Effectively the same as !delete_child_denied */
                        KAUTH_DEBUG("%p    DEFERRED - directory ACL", vcp->vp);
                default:
                        /* Effectively the same as !delete_child_denied */
                        KAUTH_DEBUG("%p    DEFERRED - directory ACL", vcp->vp);
@@ -5866,8 +6425,8 @@ vnode_authorize_callback(kauth_cred_t cred, void *idata, kauth_action_t action,
                 * find the stream and flush its cache.
                 */
                if (vnode_isnamedstream(vp) && (!vfs_authopaque(vp->v_mount))) {
                 * find the stream and flush its cache.
                 */
                if (vnode_isnamedstream(vp) && (!vfs_authopaque(vp->v_mount))) {
-                       cvp = vp->v_parent;
-                       if ((cvp != NULLVP) && (vnode_getwithref(cvp) == 0)) {
+                       cvp = vnode_getparent(vp);
+                       if (cvp != NULLVP) {
                                parent_iocount = 1;
                        } else {
                                cvp = NULL;
                                parent_iocount = 1;
                        } else {
                                cvp = NULL;
@@ -5897,8 +6456,10 @@ vnode_authorize_callback(kauth_cred_t cred, void *idata, kauth_action_t action,
 defer:
         result = vnode_authorize_callback_int(cred, idata, action, arg0, arg1, arg2, arg3);
 
 defer:
         result = vnode_authorize_callback_int(cred, idata, action, arg0, arg1, arg2, arg3);
 
-       if (result == KAUTH_RESULT_ALLOW && cvp != NULLVP)
+       if (result == KAUTH_RESULT_ALLOW && cvp != NULLVP) {
+               KAUTH_DEBUG("%p - caching action = %x", cvp, action);
                vnode_cache_authorized_action(cvp, ctx, action);
                vnode_cache_authorized_action(cvp, ctx, action);
+       }
 
 out:
        if (parent_iocount) {
 
 out:
        if (parent_iocount) {
@@ -6068,7 +6629,7 @@ vnode_authorize_callback_int(__unused kauth_cred_t unused_cred, __unused void *i
         */
        if (vnode_isnamedstream(vp) &&
            (vp->v_parent != NULL) &&
         */
        if (vnode_isnamedstream(vp) &&
            (vp->v_parent != NULL) &&
-           (vget_internal(vp->v_parent, 0, VNODE_NODEAD) == 0)) {
+           (vget_internal(vp->v_parent, 0, VNODE_NODEAD | VNODE_DRAINO) == 0)) {
                parent_ref = TRUE;
                vcp->vp = vp = vp->v_parent;
                if (VATTR_IS_SUPPORTED(&va, va_acl) && (va.va_acl != NULL))
                parent_ref = TRUE;
                vcp->vp = vp = vp->v_parent;
                if (VATTR_IS_SUPPORTED(&va, va_acl) && (va.va_acl != NULL))
@@ -6175,6 +6736,7 @@ out:
                if (VATTR_IS_SUPPORTED(&dva, va_mode) &&
                    !(dva.va_mode & (S_ISVTX))) {
                        /* OK to cache delete rights */
                if (VATTR_IS_SUPPORTED(&dva, va_mode) &&
                    !(dva.va_mode & (S_ISVTX))) {
                        /* OK to cache delete rights */
+                       KAUTH_DEBUG("%p - caching DELETE_CHILD rights", dvp);
                        vnode_cache_authorized_action(dvp, ctx, KAUTH_VNODE_DELETE_CHILD);
                }
        }
                        vnode_cache_authorized_action(dvp, ctx, KAUTH_VNODE_DELETE_CHILD);
                }
        }
@@ -6188,12 +6750,18 @@ out:
        return(KAUTH_RESULT_ALLOW);
 }
 
        return(KAUTH_RESULT_ALLOW);
 }
 
+int 
+vnode_authattr_new(vnode_t dvp, struct vnode_attr *vap, int noauth, vfs_context_t ctx)
+{
+       return vnode_authattr_new_internal(dvp, vap, noauth, NULL, ctx);
+}
+
 /*
  * Check that the attribute information in vattr can be legally applied to
  * a new file by the context.
  */
 /*
  * Check that the attribute information in vattr can be legally applied to
  * a new file by the context.
  */
-int
-vnode_authattr_new(vnode_t dvp, struct vnode_attr *vap, int noauth, vfs_context_t ctx)
+static int
+vnode_authattr_new_internal(vnode_t dvp, struct vnode_attr *vap, int noauth, uint32_t *defaulted_fieldsp, vfs_context_t ctx)
 {
        int             error;
        int             has_priv_suser, ismember, defaulted_owner, defaulted_group, defaulted_mode;
 {
        int             error;
        int             has_priv_suser, ismember, defaulted_owner, defaulted_group, defaulted_mode;
@@ -6202,6 +6770,11 @@ vnode_authattr_new(vnode_t dvp, struct vnode_attr *vap, int noauth, vfs_context_
        mount_t         dmp;
 
        error = 0;
        mount_t         dmp;
 
        error = 0;
+
+       if (defaulted_fieldsp) {
+               *defaulted_fieldsp = 0;
+       }
+
        defaulted_owner = defaulted_group = defaulted_mode = 0;
 
        /*
        defaulted_owner = defaulted_group = defaulted_mode = 0;
 
        /*
@@ -6384,6 +6957,17 @@ vnode_authattr_new(vnode_t dvp, struct vnode_attr *vap, int noauth, vfs_context_
                }
        }
 out:   
                }
        }
 out:   
+       if (defaulted_fieldsp) {
+               if (defaulted_mode) {
+                       *defaulted_fieldsp |= VATTR_PREPARE_DEFAULTED_MODE;
+               }
+               if (defaulted_group) {
+                       *defaulted_fieldsp |= VATTR_PREPARE_DEFAULTED_GID;
+               }
+               if (defaulted_owner) {
+                       *defaulted_fieldsp |= VATTR_PREPARE_DEFAULTED_UID;
+               }
+       }
        return(error);
 }
 
        return(error);
 }
 
@@ -6481,6 +7065,14 @@ vnode_authattr(vnode_t vp, struct vnode_attr *vap, kauth_action_t *actionp, vfs_
                VATTR_WANTED(&ova, va_flags);
        }
 
                VATTR_WANTED(&ova, va_flags);
        }
 
+       /*
+        * If ACLs are being changed, we need the old ACLs.
+        */
+       if (VATTR_IS_ACTIVE(vap, va_acl)) {
+               KAUTH_DEBUG("ATTR - acl changing, fetching old flags");
+               VATTR_WANTED(&ova, va_acl);
+       }
+
        /*
         * If the size is being set, make sure it's not a directory.
         */
        /*
         * If the size is being set, make sure it's not a directory.
         */
@@ -6886,7 +7478,7 @@ no_guuid_change:
                                KAUTH_DEBUG("CHMOD - adding/removing ACL entries");
                        } else if (vap->va_acl->acl_entrycount > 0) {
                                /* both ACLs have the same ACE count, said count is 1 or more, bitwise compare ACLs */
                                KAUTH_DEBUG("CHMOD - adding/removing ACL entries");
                        } else if (vap->va_acl->acl_entrycount > 0) {
                                /* both ACLs have the same ACE count, said count is 1 or more, bitwise compare ACLs */
-                               if (!memcmp(&vap->va_acl->acl_ace[0], &ova.va_acl->acl_ace[0],
+                               if (memcmp(&vap->va_acl->acl_ace[0], &ova.va_acl->acl_ace[0],
                                        sizeof(struct kauth_ace) * vap->va_acl->acl_entrycount)) {
                                        required_action |= KAUTH_VNODE_WRITE_SECURITY;
                                        KAUTH_DEBUG("CHMOD - changing ACL entries");
                                        sizeof(struct kauth_ace) * vap->va_acl->acl_entrycount)) {
                                        required_action |= KAUTH_VNODE_WRITE_SECURITY;
                                        KAUTH_DEBUG("CHMOD - changing ACL entries");
@@ -6909,29 +7501,28 @@ out:
        return(error);
 }
 
        return(error);
 }
 
+static int
+setlocklocal_callback(struct vnode *vp, __unused void *cargs)
+{
+       vnode_lock_spin(vp);
+       vp->v_flag |= VLOCKLOCAL;
+       vnode_unlock(vp);
+
+       return (VNODE_RETURNED);
+}
 
 void
 vfs_setlocklocal(mount_t mp)
 {
 
 void
 vfs_setlocklocal(mount_t mp)
 {
-       vnode_t vp;
-       
-       mount_lock(mp);
+       mount_lock_spin(mp);
        mp->mnt_kern_flag |= MNTK_LOCK_LOCAL;
        mp->mnt_kern_flag |= MNTK_LOCK_LOCAL;
+       mount_unlock(mp);
 
        /*
 
        /*
-        * We do not expect anyone to be using any vnodes at the
-        * time this routine is called. So no need for vnode locking 
+        * The number of active vnodes is expected to be
+        * very small when vfs_setlocklocal is invoked.
         */
         */
-       TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) {
-                       vp->v_flag |= VLOCKLOCAL;
-       }
-       TAILQ_FOREACH(vp, &mp->mnt_workerqueue, v_mntvnodes) {
-                       vp->v_flag |= VLOCKLOCAL;
-       }
-       TAILQ_FOREACH(vp, &mp->mnt_newvnodes, v_mntvnodes) {
-                       vp->v_flag |= VLOCKLOCAL;
-       }
-       mount_unlock(mp);
+       vnode_iterate(mp, 0, setlocklocal_callback, NULL);
 }
 
 void
 }
 
 void
@@ -6942,6 +7533,14 @@ vfs_setunmountpreflight(mount_t mp)
        mount_unlock(mp);
 }
 
        mount_unlock(mp);
 }
 
+void
+vfs_setcompoundopen(mount_t mp)
+{
+       mount_lock_spin(mp);
+       mp->mnt_compound_ops |= COMPOUND_VNOP_OPEN;
+       mount_unlock(mp);
+}
+
 void
 vn_setunionwait(vnode_t vp)
 {
 void
 vn_setunionwait(vnode_t vp)
 {
@@ -7146,13 +7745,17 @@ errno_t rmdir_remove_orphaned_appleDouble(vnode_t vp , vfs_context_t ctx, int *
                                        !((dp->d_namlen == 1 && dp->d_name[0] == '.') ||
                                            (dp->d_namlen == 2 && dp->d_name[0] == '.' && dp->d_name[1] == '.'))
                                          ) {
                                        !((dp->d_namlen == 1 && dp->d_name[0] == '.') ||
                                            (dp->d_namlen == 2 && dp->d_name[0] == '.' && dp->d_name[1] == '.'))
                                          ) {
-
-                               NDINIT(&nd_temp, DELETE, USEDVP, UIO_SYSSPACE, CAST_USER_ADDR_T(dp->d_name), ctx);
+       
+                               NDINIT(&nd_temp, DELETE, OP_UNLINK, USEDVP,
+                                      UIO_SYSSPACE, CAST_USER_ADDR_T(dp->d_name),
+                                      ctx);
                                nd_temp.ni_dvp = vp;
                                error = unlink1(ctx, &nd_temp, 0);
                                nd_temp.ni_dvp = vp;
                                error = unlink1(ctx, &nd_temp, 0);
-                               if (error && error != ENOENT) {
+
+                               if (error &&  error != ENOENT) {
                                        goto outsc;
                                }
                                        goto outsc;
                                }
+
                        }
                        cpos += dp->d_reclen;
                        dp = (struct dirent*)cpos;
                        }
                        cpos += dp->d_reclen;
                        dp = (struct dirent*)cpos;
@@ -7208,21 +7811,645 @@ lock_vnode_and_post(vnode_t vp, int kevent_num)
 #ifdef JOE_DEBUG
 static void record_vp(vnode_t vp, int count) {
         struct uthread *ut;
 #ifdef JOE_DEBUG
 static void record_vp(vnode_t vp, int count) {
         struct uthread *ut;
-        int  i;
 
 
+#if CONFIG_TRIGGERS
+       if (vp->v_resolve)
+               return;
+#endif
        if ((vp->v_flag & VSYSTEM))
                return;
 
        ut = get_bsdthread_info(current_thread());
         ut->uu_iocount += count;
 
        if ((vp->v_flag & VSYSTEM))
                return;
 
        ut = get_bsdthread_info(current_thread());
         ut->uu_iocount += count;
 
-       if (ut->uu_vpindex < 32) {
-               for (i = 0; i < ut->uu_vpindex; i++) {
-                       if (ut->uu_vps[i] == vp)
-                               return;
+       if (count == 1) {
+               if (ut->uu_vpindex < 32) {
+                       OSBacktrace((void **)&ut->uu_pcs[ut->uu_vpindex][0], 10);
+
+                       ut->uu_vps[ut->uu_vpindex] = vp;
+                       ut->uu_vpindex++;
                }
                }
-               ut->uu_vps[ut->uu_vpindex] = vp;
-               ut->uu_vpindex++;
        }
 }
 #endif
        }
 }
 #endif
+
+
+#if CONFIG_TRIGGERS
+
+#define TRIG_DEBUG 0
+
+#if TRIG_DEBUG
+#define TRIG_LOG(...) do { printf("%s: ", __FUNCTION__); printf(__VA_ARGS__); } while (0)
+#else
+#define TRIG_LOG(...)
+#endif
+
+/*
+ * Resolver result functions
+ */
+
+resolver_result_t
+vfs_resolver_result(uint32_t seq, enum resolver_status stat, int aux)
+{
+       /*
+        * |<---   32   --->|<---  28  --->|<- 4 ->|
+        *      sequence        auxiliary    status
+        */
+       return (((uint64_t)seq) << 32) |                
+              (((uint64_t)(aux & 0x0fffffff)) << 4) |  
+              (uint64_t)(stat & 0x0000000F);   
+}
+
+enum resolver_status
+vfs_resolver_status(resolver_result_t result)
+{
+       /* lower 4 bits is status */
+       return (result & 0x0000000F);
+}
+
+uint32_t
+vfs_resolver_sequence(resolver_result_t result)
+{
+       /* upper 32 bits is sequence */
+       return (uint32_t)(result >> 32);
+}
+
+int
+vfs_resolver_auxiliary(resolver_result_t result)
+{
+       /* 28 bits of auxiliary */
+       return (int)(((uint32_t)(result & 0xFFFFFFF0)) >> 4);
+}
+
+/*
+ * SPI
+ * Call in for resolvers to update vnode trigger state
+ */
+int
+vnode_trigger_update(vnode_t vp, resolver_result_t result)
+{
+       vnode_resolve_t rp;
+       uint32_t seq;
+       enum resolver_status stat;
+
+       if (vp->v_resolve == NULL) {
+               return (EINVAL);
+       }
+
+       stat = vfs_resolver_status(result);
+       seq = vfs_resolver_sequence(result);
+
+       if ((stat != RESOLVER_RESOLVED) && (stat != RESOLVER_UNRESOLVED)) {
+               return (EINVAL);
+       }
+
+       rp = vp->v_resolve;
+       lck_mtx_lock(&rp->vr_lock);
+
+       if (seq > rp->vr_lastseq) {
+               if (stat == RESOLVER_RESOLVED)
+                       rp->vr_flags |= VNT_RESOLVED;
+               else
+                       rp->vr_flags &= ~VNT_RESOLVED;
+
+               rp->vr_lastseq = seq;
+       }
+
+       lck_mtx_unlock(&rp->vr_lock);
+
+       return (0);
+}
+
+static int
+vnode_resolver_attach(vnode_t vp, vnode_resolve_t rp, boolean_t ref)
+{
+       int error;
+
+       vnode_lock_spin(vp);
+       if (vp->v_resolve != NULL) {
+               vnode_unlock(vp);
+               return EINVAL;
+       } else {
+               vp->v_resolve = rp;
+       }
+       vnode_unlock(vp);
+       
+       if (ref) {
+               error = vnode_ref_ext(vp, O_EVTONLY, VNODE_REF_FORCE);
+               if (error != 0) {
+                       panic("VNODE_REF_FORCE didn't help...");
+               }
+       }
+
+       return 0;
+}
+
+/*
+ * VFS internal interfaces for vnode triggers
+ *
+ * vnode must already have an io count on entry
+ * v_resolve is stable when io count is non-zero
+ */
+static int
+vnode_resolver_create(mount_t mp, vnode_t vp, struct vnode_trigger_param *tinfo, boolean_t external)
+{
+       vnode_resolve_t rp;
+       int result;
+       char byte;
+
+#if 1
+       /* minimum pointer test (debugging) */
+       if (tinfo->vnt_data)
+               byte = *((char *)tinfo->vnt_data);
+#endif
+       MALLOC(rp, vnode_resolve_t, sizeof(*rp), M_TEMP, M_WAITOK);
+       if (rp == NULL)
+               return (ENOMEM);
+
+       lck_mtx_init(&rp->vr_lock, trigger_vnode_lck_grp, trigger_vnode_lck_attr);
+
+       rp->vr_resolve_func = tinfo->vnt_resolve_func;
+       rp->vr_unresolve_func = tinfo->vnt_unresolve_func;
+       rp->vr_rearm_func = tinfo->vnt_rearm_func;
+       rp->vr_reclaim_func = tinfo->vnt_reclaim_func;
+       rp->vr_data = tinfo->vnt_data;
+       rp->vr_lastseq = 0;
+       rp->vr_flags = tinfo->vnt_flags & VNT_VALID_MASK;
+       if (external) {
+               rp->vr_flags |= VNT_EXTERNAL;
+       }
+
+       result = vnode_resolver_attach(vp, rp, external);
+       if (result != 0) {
+               goto out;
+       }
+
+       if (mp) {
+               OSAddAtomic(1, &mp->mnt_numtriggers);
+       }
+
+       return (result);
+
+out:
+       FREE(rp, M_TEMP);
+       return result;
+}
+
+static void
+vnode_resolver_release(vnode_resolve_t rp)
+{
+       /*
+        * Give them a chance to free any private data
+        */
+       if (rp->vr_data && rp->vr_reclaim_func) {
+               rp->vr_reclaim_func(NULLVP, rp->vr_data);
+       }
+
+       lck_mtx_destroy(&rp->vr_lock, trigger_vnode_lck_grp);
+       FREE(rp, M_TEMP);
+
+}
+
+/* Called after the vnode has been drained */
+static void
+vnode_resolver_detach(vnode_t vp)
+{
+       vnode_resolve_t rp;
+       mount_t mp;
+
+       mp = vnode_mount(vp);
+
+       vnode_lock(vp);
+       rp = vp->v_resolve;
+       vp->v_resolve = NULL;
+       vnode_unlock(vp);
+
+       if ((rp->vr_flags & VNT_EXTERNAL) != 0) {
+               vnode_rele_ext(vp, O_EVTONLY, 1);
+       } 
+
+       vnode_resolver_release(rp);
+       
+       /* Keep count of active trigger vnodes per mount */
+       OSAddAtomic(-1, &mp->mnt_numtriggers);  
+}
+
+/*
+ * Pathname operations that don't trigger a mount for trigger vnodes
+ */
+static const u_int64_t ignorable_pathops_mask =
+       1LL << OP_MOUNT |
+       1LL << OP_UNMOUNT |
+       1LL << OP_STATFS |
+       1LL << OP_ACCESS |
+       1LL << OP_GETATTR |
+       1LL << OP_LISTXATTR;
+
+int
+vfs_istraditionaltrigger(enum path_operation op, const struct componentname *cnp)
+{
+       if (cnp->cn_flags & ISLASTCN)
+               return ((1LL << op) & ignorable_pathops_mask) == 0;
+       else
+               return (1);
+}
+
+__private_extern__
+void
+vnode_trigger_rearm(vnode_t vp, vfs_context_t ctx)
+{
+       vnode_resolve_t rp;
+       resolver_result_t result;
+       enum resolver_status status;
+       uint32_t seq;
+
+       if ((vp->v_resolve == NULL) ||
+           (vp->v_resolve->vr_rearm_func == NULL) ||
+           (vp->v_resolve->vr_flags & VNT_AUTO_REARM) == 0) {
+               return;
+       }
+
+       rp = vp->v_resolve;
+       lck_mtx_lock(&rp->vr_lock);
+
+       /*
+        * Check if VFS initiated this unmount. If so, we'll catch it after the unresolve completes.
+        */
+       if (rp->vr_flags & VNT_VFS_UNMOUNTED) {
+               lck_mtx_unlock(&rp->vr_lock);
+               return;
+       }
+
+       /* Check if this vnode is already armed */
+       if ((rp->vr_flags & VNT_RESOLVED) == 0) {
+               lck_mtx_unlock(&rp->vr_lock);
+               return;
+       }
+
+       lck_mtx_unlock(&rp->vr_lock);
+
+       result = rp->vr_rearm_func(vp, 0, rp->vr_data, ctx);
+       status = vfs_resolver_status(result);
+       seq = vfs_resolver_sequence(result);
+
+       lck_mtx_lock(&rp->vr_lock);
+       if (seq > rp->vr_lastseq) {
+               if (status == RESOLVER_UNRESOLVED)
+                       rp->vr_flags &= ~VNT_RESOLVED;
+               rp->vr_lastseq = seq;
+       }
+       lck_mtx_unlock(&rp->vr_lock);
+}
+
+__private_extern__
+int
+vnode_trigger_resolve(vnode_t vp, struct nameidata *ndp, vfs_context_t ctx)
+{
+       vnode_resolve_t rp;
+       enum path_operation op;
+       resolver_result_t result;
+       enum resolver_status status;
+       uint32_t seq;
+
+       /* Only trigger on topmost vnodes */
+       if ((vp->v_resolve == NULL) ||
+           (vp->v_resolve->vr_resolve_func == NULL) ||
+           (vp->v_mountedhere != NULL)) {
+               return (0);
+       }
+
+       rp = vp->v_resolve;
+       lck_mtx_lock(&rp->vr_lock);
+
+       /* Check if this vnode is already resolved */
+       if (rp->vr_flags & VNT_RESOLVED) {
+               lck_mtx_unlock(&rp->vr_lock);
+               return (0);
+       }
+
+       lck_mtx_unlock(&rp->vr_lock);
+
+       /*
+        * XXX
+        * assumes that resolver will not access this trigger vnode (otherwise the kernel will deadlock)
+        * is there anyway to know this???
+        * there can also be other legitimate lookups in parallel
+        *
+        * XXX - should we call this on a separate thread with a timeout?
+        * 
+        * XXX - should we use ISLASTCN to pick the op value???  Perhaps only leafs should
+        * get the richer set and non-leafs should get generic OP_LOOKUP?  TBD
+        */
+       op = (ndp->ni_op < OP_MAXOP) ? ndp->ni_op: OP_LOOKUP;
+
+       result = rp->vr_resolve_func(vp, &ndp->ni_cnd, op, 0, rp->vr_data, ctx);
+       status = vfs_resolver_status(result);
+       seq = vfs_resolver_sequence(result);
+
+       lck_mtx_lock(&rp->vr_lock);
+       if (seq > rp->vr_lastseq) {
+               if (status == RESOLVER_RESOLVED)
+                       rp->vr_flags |= VNT_RESOLVED;
+               rp->vr_lastseq = seq;
+       }
+       lck_mtx_unlock(&rp->vr_lock);
+
+       /* On resolver errors, propagate the error back up */
+       return (status == RESOLVER_ERROR ? vfs_resolver_auxiliary(result) : 0);
+}
+
+static int
+vnode_trigger_unresolve(vnode_t vp, int flags, vfs_context_t ctx)
+{
+       vnode_resolve_t rp;
+       resolver_result_t result;
+       enum resolver_status status;
+       uint32_t seq;
+
+       if ((vp->v_resolve == NULL) || (vp->v_resolve->vr_unresolve_func == NULL)) {
+               return (0);
+       }
+
+       rp = vp->v_resolve;
+       lck_mtx_lock(&rp->vr_lock);
+
+       /* Check if this vnode is already resolved */
+       if ((rp->vr_flags & VNT_RESOLVED) == 0) {
+               printf("vnode_trigger_unresolve: not currently resolved\n");
+               lck_mtx_unlock(&rp->vr_lock);
+               return (0);
+       }
+
+       rp->vr_flags |= VNT_VFS_UNMOUNTED;
+
+       lck_mtx_unlock(&rp->vr_lock);
+
+       /*
+        * XXX
+        * assumes that resolver will not access this trigger vnode (otherwise the kernel will deadlock)
+        * there can also be other legitimate lookups in parallel
+        *
+        * XXX - should we call this on a separate thread with a timeout?
+        */
+
+       result = rp->vr_unresolve_func(vp, flags, rp->vr_data, ctx);
+       status = vfs_resolver_status(result);
+       seq = vfs_resolver_sequence(result);
+
+       lck_mtx_lock(&rp->vr_lock);
+       if (seq > rp->vr_lastseq) {
+               if (status == RESOLVER_UNRESOLVED)
+                       rp->vr_flags &= ~VNT_RESOLVED;
+               rp->vr_lastseq = seq;
+       }
+       rp->vr_flags &= ~VNT_VFS_UNMOUNTED;
+       lck_mtx_unlock(&rp->vr_lock);
+
+       /* On resolver errors, propagate the error back up */
+       return (status == RESOLVER_ERROR ? vfs_resolver_auxiliary(result) : 0);
+}
+
+static int
+triggerisdescendant(mount_t mp, mount_t rmp)
+{
+       int match = FALSE;
+
+       /*
+        * walk up vnode covered chain looking for a match
+        */
+       name_cache_lock_shared();
+
+       while (1) {
+               vnode_t vp;
+
+               /* did we encounter "/" ? */
+               if (mp->mnt_flag & MNT_ROOTFS)
+                       break;
+
+               vp = mp->mnt_vnodecovered;
+               if (vp == NULLVP)
+                       break;
+
+               mp = vp->v_mount;
+               if (mp == rmp) {
+                       match = TRUE;
+                       break;
+               }
+       }
+
+       name_cache_unlock();
+
+       return (match);
+}
+
+struct trigger_unmount_info {
+       vfs_context_t   ctx;
+       mount_t         top_mp;
+       vnode_t         trigger_vp;
+       mount_t         trigger_mp;
+       uint32_t        trigger_vid;
+       int             flags;
+};
+
+static int
+trigger_unmount_callback(mount_t mp, void * arg)
+{
+       struct trigger_unmount_info * infop = (struct trigger_unmount_info *)arg;
+       boolean_t mountedtrigger = FALSE;
+
+       /*
+        * When we encounter the top level mount we're done
+        */
+       if (mp == infop->top_mp)
+               return (VFS_RETURNED_DONE);
+
+       if ((mp->mnt_vnodecovered == NULL) ||
+           (vnode_getwithref(mp->mnt_vnodecovered) != 0)) {
+               return (VFS_RETURNED);
+       }
+
+       if ((mp->mnt_vnodecovered->v_mountedhere == mp) &&
+           (mp->mnt_vnodecovered->v_resolve != NULL) &&
+           (mp->mnt_vnodecovered->v_resolve->vr_flags & VNT_RESOLVED)) {
+               mountedtrigger = TRUE;
+       }
+       vnode_put(mp->mnt_vnodecovered);
+
+       /*
+        * When we encounter a mounted trigger, check if its under the top level mount
+        */
+       if ( !mountedtrigger || !triggerisdescendant(mp, infop->top_mp) )
+               return (VFS_RETURNED);
+
+       /*
+        * Process any pending nested mount (now that its not referenced)
+        */
+       if ((infop->trigger_vp != NULLVP) &&
+           (vnode_getwithvid(infop->trigger_vp, infop->trigger_vid) == 0)) {
+               vnode_t vp = infop->trigger_vp;
+               int error;
+
+               infop->trigger_vp = NULLVP;
+               
+               if (mp == vp->v_mountedhere) {
+                       vnode_put(vp);
+                       printf("trigger_unmount_callback: unexpected match '%s'\n",
+                               mp->mnt_vfsstat.f_mntonname);
+                       return (VFS_RETURNED);
+               }
+               if (infop->trigger_mp != vp->v_mountedhere) {
+                       vnode_put(vp);
+                       printf("trigger_unmount_callback: trigger mnt changed! (%p != %p)\n",
+                               infop->trigger_mp, vp->v_mountedhere);
+                       goto savenext;
+               }
+
+               error = vnode_trigger_unresolve(vp, infop->flags, infop->ctx);
+               vnode_put(vp);
+               if (error) {
+                       printf("unresolving: '%s', err %d\n",
+                               vp->v_mountedhere ? vp->v_mountedhere->mnt_vfsstat.f_mntonname :
+                               "???", error);
+                       return (VFS_RETURNED_DONE); /* stop iteration on errors */
+               }
+       }
+savenext:
+       /*
+        * We can't call resolver here since we hold a mount iter
+        * ref on mp so save its covered vp for later processing
+        */
+       infop->trigger_vp = mp->mnt_vnodecovered;
+       if ((infop->trigger_vp != NULLVP) &&
+           (vnode_getwithref(infop->trigger_vp) == 0)) {
+               if (infop->trigger_vp->v_mountedhere == mp) {
+                       infop->trigger_vid = infop->trigger_vp->v_id;
+                       infop->trigger_mp = mp;
+               }
+               vnode_put(infop->trigger_vp);
+       }
+
+       return (VFS_RETURNED);
+}
+
+/*
+ * Attempt to unmount any trigger mounts nested underneath a mount.
+ * This is a best effort attempt and no retries are performed here.
+ *
+ * Note: mp->mnt_rwlock is held exclusively on entry (so be carefull)
+ */
+__private_extern__
+void
+vfs_nested_trigger_unmounts(mount_t mp, int flags, vfs_context_t ctx)
+{
+       struct trigger_unmount_info info;
+
+       /* Must have trigger vnodes */
+       if (mp->mnt_numtriggers == 0) {
+               return;
+       }
+       /* Avoid recursive requests (by checking covered vnode) */
+       if ((mp->mnt_vnodecovered != NULL) &&
+           (vnode_getwithref(mp->mnt_vnodecovered) == 0)) {
+               boolean_t recursive = FALSE;
+
+               if ((mp->mnt_vnodecovered->v_mountedhere == mp) &&
+                   (mp->mnt_vnodecovered->v_resolve != NULL) &&
+                   (mp->mnt_vnodecovered->v_resolve->vr_flags & VNT_VFS_UNMOUNTED)) {
+                       recursive = TRUE;
+               }
+               vnode_put(mp->mnt_vnodecovered);
+               if (recursive)
+                       return;
+       }
+
+       /*
+        * Attempt to unmount any nested trigger mounts (best effort)
+        */
+       info.ctx = ctx;
+       info.top_mp = mp;
+       info.trigger_vp = NULLVP;
+       info.trigger_vid = 0;
+       info.trigger_mp = NULL;
+       info.flags = flags;
+
+       (void) vfs_iterate(VFS_ITERATE_TAIL_FIRST, trigger_unmount_callback, &info);
+
+       /*
+        * Process remaining nested mount (now that its not referenced)
+        */
+       if ((info.trigger_vp != NULLVP) &&
+           (vnode_getwithvid(info.trigger_vp, info.trigger_vid) == 0)) {
+               vnode_t vp = info.trigger_vp;
+
+               if (info.trigger_mp == vp->v_mountedhere) {
+                       (void) vnode_trigger_unresolve(vp, flags, ctx);
+               }
+               vnode_put(vp);
+       }
+}
+
+int
+vfs_addtrigger(mount_t mp, const char *relpath, struct vnode_trigger_info *vtip, vfs_context_t ctx)
+{
+       struct nameidata nd;
+       int res;
+       vnode_t rvp, vp;
+       struct vnode_trigger_param vtp;
+       
+       /* 
+        * Must be called for trigger callback, wherein rwlock is held 
+        */
+       lck_rw_assert(&mp->mnt_rwlock, LCK_RW_ASSERT_HELD);
+
+       TRIG_LOG("Adding trigger at %s\n", relpath);
+       TRIG_LOG("Trying VFS_ROOT\n");
+
+       /* 
+        * We do a lookup starting at the root of the mountpoint, unwilling
+        * to cross into other mountpoints.
+        */
+       res = VFS_ROOT(mp, &rvp, ctx);
+       if (res != 0) {
+               goto out;
+       }
+
+       TRIG_LOG("Trying namei\n");
+
+       NDINIT(&nd, LOOKUP, OP_LOOKUP, USEDVP | NOCROSSMOUNT | FOLLOW, UIO_SYSSPACE,
+               CAST_USER_ADDR_T(relpath), ctx);
+       nd.ni_dvp = rvp;
+       res = namei(&nd);
+       if (res != 0) {
+               vnode_put(rvp);
+               goto out;
+       }
+       
+       vp = nd.ni_vp;
+       nameidone(&nd);
+       vnode_put(rvp);
+
+       TRIG_LOG("Trying vnode_resolver_create()\n");
+
+       /* 
+        * Set up blob.  vnode_create() takes a larger structure
+        * with creation info, and we needed something different
+        * for this case.  One needs to win, or we need to munge both;
+        * vnode_create() wins.
+        */
+       bzero(&vtp, sizeof(vtp));
+       vtp.vnt_resolve_func = vtip->vti_resolve_func;
+       vtp.vnt_unresolve_func = vtip->vti_unresolve_func;
+       vtp.vnt_rearm_func = vtip->vti_rearm_func;
+       vtp.vnt_reclaim_func = vtip->vti_reclaim_func;
+       vtp.vnt_reclaim_func = vtip->vti_reclaim_func;
+       vtp.vnt_data = vtip->vti_data;
+       vtp.vnt_flags = vtip->vti_flags;
+
+       res = vnode_resolver_create(mp, vp, &vtp, TRUE);
+       vnode_put(vp);
+out:
+       TRIG_LOG("Returning %d\n", res);
+       return res;
+}
+
+#endif /* CONFIG_TRIGGERS */